OSCR

Age-Related Differences in Resting-State Functional Connectivity Predict Specific Patterns of Speech Disfluency.

Code ↔ Paper

3 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 3 matches
  1. [1] § METHODS › Predicting Disfluency From Age, EF, and Network Segregation ↔ Analysis Files/RSFC_stats_analysis_OSF.Rmd, lines 1130–1157 · score 0.57 · Johnson Neyman, simple slopes, intervals, interactively, predicted, Education
  2. [2] § METHODS › Participant Demographics ↔ Analysis Files/RSFC_stats_analysis_OSF.Rmd, lines 120–170 · score 0.55 · Cook, distance, threshold, outliers, MMSE, MoCA
  3. [3] § RESULTS › Network Segregation Predicts Disfluency ↔ Analysis Files/RSFC_stats_analysis_OSF.Rmd, lines 1130–1157 · score 0.52 · Johnson Neyman, simple slopes, DMN segregation, younger, interaction, older

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

R Markdown · 1,813 lines · 56 KB · no license · 3 matches

  1. ---
  2. title: "RSFC_stats_analysis"
  3. output: html_document
  4. date: "2025-04-25"
  5. ---
  7. ```{r load libraries}
  8. library ("readr")
  9. library("tidyverse")
  10. library("ggplot2")
  11. library("dplyr")
  12. library("broom")
  13. library("knitr")
  14. library("psych")
  15. library("kableExtra")
  16. ```
  17. This script runs the regression and mediation analysis for RSFC data for the DMN, MD, and Language networks and the Speech disfluency data for a dataset comprising thress cohorts "PND", "MP", and "PND2".
  19. Master_data_updated.csv contains all behavioral data including Fluency (for tasks "Speech" and "Where"), Stroop, moca/mmse, etc.
  21. All rsfc files contain subject measures of resting state functional connectivity: segregation, within, and between network connectivity
  23. Note on "discourse type" aka speech elcitation task:
  24. For MP and PND2, this script calculates averages collapsed across "Speech" and "Where" (Open-ended prompt and "Frog Where Are You?" respectively)
  26. ```{r load data}
  27. bhv_data <- read_csv(
  28. "master_data_updated.csv",
  29. col_types = cols(sex = col_character())
  30. )
  31. rsfc_data_PND <- read_csv("PND_subject_measures.csv")
  32. rsfc_data_MP <- read_csv("MP_subject_measures.csv")
  33. rsfc_data_PND2 <- read_csv("PND2_subject_measures.csv")
  35. #view(bhv_data)
  36. #view(rsfc_data_PND)
  37. #view(rsfc_data_MP)
  38. #view(rsfc_data_PND2)
  39. ```
  40. #Data Wrangling Steps
  41. Here we are collapsing Where and Speech for MP and PND2, merging the Speech data with the RSFC data, and getting a count of the sample before and after exclusion criteria
  42. ```{r clean behavioral data}
  43. #Collapse Where and Speech data for MP and PND2
  44. #Define disfluency columns to average
  45. cols_to_average <- c(
  46. "mor_words", "perc_prolongation", "perc_wwr", "perc_total_repetitions", "perc_mono_wwr",
  47. "perc_phrase_repetitions", "perc_word_revisions", "perc_phrase_revisions", "perc_total_revisions",
  48. "perc_filled_pauses", "perc_pauses", "perc_typical_disfluencies", "perc_stutter_disfluencies",
  49. "perc_total_disfluencies", "freq_types", "freq_tokens", "freq_ttr", "total_trad_fps",
  50. "total_fillers", "percentage_fillers", "percentage_trad_fps", "duration", "speech_rate"
  51. )
  53. #Collapse row pairs for MP
  54. collapsed_MP <- bhv_data %>%
  55. filter(project == "MP") %>%
  56. group_by(subject) %>%
  57. summarise(across(all_of(cols_to_average), mean, na.rm = TRUE), .groups = "drop") %>%
  58. left_join(
  59. bhv_data %>% filter(project == "MP") %>% group_by(subject) %>% slice(1) %>%
  60. select(-all_of(cols_to_average), -discourse_type),
  61. by = "subject"
  62. ) %>%
  63. mutate(discourse_type = "BOTH") %>% #change discourse type label
  64. relocate(project)
  66. #Collapse row pairs for PND2
  67. collapsed_PND2 <- bhv_data %>%
  68. filter(project == "PND2") %>%
  69. group_by(subject) %>%
  70. summarise(across(all_of(cols_to_average), mean, na.rm = TRUE), .groups = "drop") %>%
  71. left_join(
  72. bhv_data %>% filter(project == "PND2") %>% group_by(subject) %>% slice(1) %>%
  73. select(-all_of(cols_to_average), -discourse_type),
  74. by = "subject"
  75. ) %>%
  76. mutate(discourse_type = "BOTH") %>% #change discourse type label
  77. relocate(project)
  79. #Keep PND
  80. PND_unchanged <- bhv_data %>%
  81. filter(project == "PND")
  83. #Bind all three dataframes
  84. final_bhv_df <- bind_rows(PND_unchanged, collapsed_MP, collapsed_PND2)
  86. View(final_bhv_df)
  87. ```
  88. ```{r Merge Behavioral data w RSFC data}
  89. #Merge all Fluency data with RSFC data
  90. rsfc_data_PND <- rsfc_data_PND %>%
  91. rename(subject = Subject_ID) %>%
  92. mutate(project = "PND")
  94. rsfc_data_MP <- rsfc_data_MP %>%
  95. rename(subject = Subject_ID) %>%
  96. mutate(project = "MP")
  98. rsfc_data_PND2 <- rsfc_data_PND2 %>%
  99. rename(subject = Subject_ID) %>%
  100. mutate(project = "PND2")
  102. #Combine all RSFC datasets
  103. rsfc_data_all <- bind_rows(rsfc_data_PND, rsfc_data_MP, rsfc_data_PND2)
  105. #Merge RSFC with behavioral data
  106. merged_initial <- final_bhv_df %>%
  107. left_join(rsfc_data_all, by = c("subject", "project"))
  109. # Get initial sample size BEFORE exclusions
  110. initial_n <- nrow(merged_initial)
  112. cat("Initial sample size (before exclusions):", initial_n, "\n")
  114. view(merged_initial)
  116. range_age <- range(merged_initial$age, na.rm = TRUE)
  117. cat("Youngest participant age:", range_age[1], "\n")
  118. cat("Oldest participant age:", range_age[2], "\n")
  119. ```
  120. ```{r Apply Exclusion criteria}
  121. # Exclude based on moca and mmse
  122. excluded_cog <- merged_initial %>%
  123. filter((!is.na(mmse) & mmse < 24) | (!is.na(moca_score) & moca_score < 26))
  125. tmp_step1 <- merged_initial %>%
  126. filter(is.na(mmse) | mmse >= 24) %>%
  127. filter(is.na(moca_score) | moca_score >= 26)
  129. # Exclude based on missing key variables
  130. key_vars <- c("stroop_effect", "perc_total_disfluencies",
  131. "Lang_LH_within", "DMN_within", "MD_within",
  132. "Lang_to_DMN", "Lang_to_MD", "DMN_to_MD")
  134. excluded_missing <- tmp_step1 %>%
  135. filter(if_any(all_of(key_vars), is.na))
  137. tmp_step2 <- tmp_step1 %>%
  138. filter(if_all(all_of(key_vars), ~ !is.na(.)))
  140. # Exclude based on Cook’s Distance
  141. model_cook <- lm(perc_total_disfluencies ~ age + education, data = tmp_step2)
  142. cooks_d <- cooks.distance(model_cook)
  143. threshold <- 4 / nrow(tmp_step2)
  145. excluded_outliers <- tmp_step2[which(cooks_d > threshold), ]
  147. merged_data <- tmp_step2[which(cooks_d <= threshold), ]
  149. # Summary of Exclusions
  150. cat("Initial sample size:", initial_n, "\n")
  151. cat("Excluded for cognitive criteria:", nrow(excluded_cog), "\n")
  152. cat("Excluded for missing key variables:", nrow(excluded_missing), "\n")
  153. cat("Excluded as outliers based on Cook’s Distance:", nrow(excluded_outliers), "\n")
  154. cat("Final analytic sample:", nrow(merged_data), "\n\n")
  156. # Use Participant IDs to see who was excluded and why
  157. #excluded_cog %>%
  158. #select(subject, project) %>%
  159. #print(n = Inf)
  161. #excluded_missing %>%
  162. #select(subject, project) %>%
  163. #print(n = Inf)
  165. #excluded_outliers %>%
  166. #select(subject, project) %>%
  167. #print(n = Inf)
  169. view(merged_data)
  170. ```
  172. #Quality Check data distrubutions
  173. ```{r QC key variables}
  174. # Select key numeric variables for QC
  175. qc_vars <- c("age", "education", "stroop_effect",
  176. "perc_total_disfluencies", "perc_total_repetitions",
  177. "perc_filled_pauses", "perc_pauses", "perc_prolongation",
  178. "perc_total_revisions")
  180. # Run descriptive statistics (includes skew, kurtosis)
  181. describe(merged_data[qc_vars])
  182. ```
  183. ```{r QC plots}
  184. qc_vars <- c(
  185. "age",
  186. "education",
  187. "stroop_effect",
  188. "perc_total_disfluencies",
  189. "perc_total_repetitions",
  190. "perc_filled_pauses",
  191. "perc_pauses",
  192. "perc_prolangation",
  193. "perc_total_revisions"
  194. )
  196. # Loop through variables
  197. for (var in qc_vars) {
  198. data_vec <- merged_data[[var]]
  200. # Skip non-numeric just in case
  201. if (is.numeric(data_vec)) {
  202. # Set up side-by-side plots
  203. par(mfrow = c(1, 2))
  205. # Histogram
  206. hist(data_vec,
  207. main = paste("Histogram of", var),
  208. xlab = var,
  209. col = "lightgray",
  210. border = "white")
  212. # Q-Q plot
  213. qqnorm(data_vec, main = paste("Q-Q Plot of", var))
  214. qqline(data_vec, col = "red", lwd = 2)
  216. # Pause between variables to review each set
  217. readline(prompt = paste("Press [Enter] to continue to next variable:", var))
  218. }
  219. }
  221. # Reset plotting layout back to default
  222. par(mfrow = c(1, 1))
  223. ```
  224. #Get Descriptives
  225. ```{r Desciptives}
  226. merged_data %>%
  227. summarise(
  228. N = n(),
  229. mean_age = mean(age, na.rm = TRUE),
  230. sd_age = sd(age, na.rm = TRUE),
  231. min_age = min(age, na.rm = TRUE),
  232. max_age = max(age, na.rm = TRUE),
  233. n_female = sum(sex == "F", na.rm = TRUE),
  234. mean_education = mean(education, na.rm = TRUE),
  235. min_edu = min(education, na.rm = TRUE),
  236. max_edu = max(education, na.rm = TRUE),
  237. sd_education = sd(education, na.rm = TRUE)
  238. )
  239. ```
  240. #Get cog/ling battery performance
  241. Note that here we are calculating means only for mmse, moca, stroop interference effect, forward digit span, backward digit span, pitt_acc (reading span accuracy), and pitt_pu (reading span partial unit score), and VF semantic category total tokens. Descriptives for the other tasks e.g., WAISS, speed of processing, etc. were calculated in an additional .rmd called RSFC_cog_battery_means.rmd.
  242. ```{r cognitive performance means}
  243. cog_vars <- c(
  244. "mmse", "moca_score", "stroop_effect", "forward_digit", "backward_digit",
  245. "pitt_acc", "pitt_pu", "phon_total", "sem_total")
  247. #summary table for mean, SD, and age beta
  248. summary_stats <- map_dfr(cog_vars, function(var) {
  249. this_data <- merged_data[[var]]
  250. model <- lm(reformulate("age", response = var), data = merged_data)
  252. tibble(
  253. variable = var,
  254. mean = mean(this_data, na.rm = TRUE),
  255. sd = sd(this_data, na.rm = TRUE),
  256. min = min(this_data, na.rm = TRUE),
  257. max = max(this_data, na.rm = TRUE),
  258. beta_age = coef(model)["age"]
  259. )
  260. })
  262. summary_stats
  263. ```
  265. #Start of LM analyses
  266. He we begin the LM analyses. We ran seperate models to test if Age, EF, and all RSFC measures predicted any of the 6 disfluency subtypes.
  268. We started first with simple Age effects for disfluency, RSFC, and EF (Stroop) performance.
  270. ## Age effects
  271. ```{r Age effects for disfluency}
  272. #first make sure 'project' and is a factor to have it as a covariate
  273. merged_data$project <- factor(merged_data$project)
  275. disfluency_vars <- c(
  276. "perc_total_disfluencies",
  277. "perc_total_repetitions",
  278. "perc_filled_pauses",
  279. "perc_pauses",
  280. "perc_total_revisions",
  281. "perc_prolongation"
  282. )
  284. age_effects_disfluency <- data.frame()
  286. for (var in disfluency_vars) {
  287. # Fit the model controlling for age education, and project
  288. frm <- as.formula(paste(var, "~ age + education + project"))
  289. model_disfluency <- lm(frm, data = merged_data)
  291. #Get age coefficient and CI
  292. s <- summary(model_disfluency)
  293. age_coef <- coef(s)["age", ]
  294. age_ci <- confint(model_disfluency)["age", ] # 95% CI
  296. #Build a results row including CI
  297. result_row <- data.frame(
  298. outcome = var,
  299. estimate = age_coef["Estimate"],
  300. std.error = age_coef["Std. Error"],
  301. statistic = age_coef["t value"],
  302. p.value = age_coef["Pr(>|t|)"],
  303. CI_lower = age_ci[1],
  304. CI_upper = age_ci[2]
  305. )
  307. age_effects_disfluency <- rbind(age_effects_disfluency, result_row)
  308. }
  310. #Apply fdr correction across the six tests
  311. age_effects_disfluency$adj.p <- p.adjust(age_effects_disfluency$p.value,
  312. method = "fdr")
  314. print(age_effects_disfluency)
  315. ```
  316. ```{r Age on disfluency_results_table}
  317. age_effects_disfluency %>%
  318. kable(
  319. format = ifelse(knitr::is_latex_output(), "latex", "html"),
  320. digits = 3,
  321. caption = "Age Effects on Disfluency Measures (fdr-adjusted)"
  322. ) %>%
  323. kable_styling(full_width = FALSE)
  325. ```
  326. ```{r Age effects for RSFC}
  327. rsfc_vars <- c(
  328. "Lang_LH_within", "DMN_within", "MD_within",
  329. "Lang_to_DMN", "Lang_to_MD",
  330. "Lang_LH_segregation", "DMN_segregation", "MD_segregation"
  331. )
  333. #Create an empty data frame to store results
  334. age_effects_rsfc <- data.frame()
  336. #Loop through each RSFC variable
  337. for (var in rsfc_vars) {
  339. #Fit LM
  340. model_RSFC <- lm(as.formula(paste(var, "~ age + education + project")), data = merged_data)
  342. #Extract summary for the age predictor
  343. model_summary <- summary(model_RSFC)
  344. age_coef <- coef(model_summary)["age", ]
  346. #Store results
  347. result_row <- data.frame(
  348. outcome = var,
  349. estimate = age_coef["Estimate"],
  350. std.error = age_coef["Std. Error"],
  351. statistic = age_coef["t value"],
  352. p.value = age_coef["Pr(>|t|)"]
  353. )
  354. #Get age coefficient and CI
  355. s <- summary(model_RSFC)
  356. age_coef <- coef(s)["age", ]
  357. age_ci <- confint(model_RSFC)["age", ] # 95% CI
  359. #Build a results row including CI
  360. result_row <- data.frame(
  361. outcome = var,
  362. estimate = age_coef["Estimate"],
  363. std.error = age_coef["Std. Error"],
  364. statistic = age_coef["t value"],
  365. p.value = age_coef["Pr(>|t|)"],
  366. CI_lower = age_ci[1],
  367. CI_upper = age_ci[2]
  368. )
  370. age_effects_rsfc <- rbind(age_effects_rsfc, result_row)
  371. }
  372. ```
  373. ```{r Age effects just on network segregation}
  374. seg_vars <- c("Lang_LH_segregation", "DMN_segregation", "MD_segregation")
  376. # lm age models
  377. age_seg <- map_dfr(seg_vars, function(var) {
  378. m <- lm(as.formula(paste(var, "~ age + education + project")), data = merged_data)
  379. est <- coef(summary(m))["age", "Estimate"]
  380. se <- coef(summary(m))["age", "Std. Error"]
  381. t <- coef(summary(m))["age", "t value"]
  382. p <- coef(summary(m))["age", "Pr(>|t|)"]
  383. ci <- confint(m)["age", ]
  384. tibble(
  385. network = var,
  386. estimate = est,
  387. std.error = se,
  388. t.value = t,
  389. p.value = p,
  390. CI_lower = ci[1],
  391. CI_upper = ci[2]
  392. )
  393. })
  395. # FDR-correct across the three p-values
  396. age_seg <- age_seg %>%
  397. mutate(q_fdr = p.adjust(p.value, method = "fdr"))
  399. # Print the table
  400. age_seg %>%
  401. kable(digits = 3, caption = "Age Effects on Network Segregation (FDR-adjusted)") %>%
  402. kable_styling(full_width = FALSE)
  403. ```
  404. ```{r Age effects for RSFC}
  405. #Apply fdr correction across the 8 tests
  406. age_effects_rsfc$adj.p <- p.adjust(age_effects_rsfc$p.value,
  407. method = "fdr")
  409. print(age_effects_rsfc)
  411. age_effects_rsfc %>%
  412. kable(
  413. format = ifelse(knitr::is_latex_output(), "latex", "html"),
  414. digits = 3,
  415. caption = "Age Effects on RSFC Measures (fdr-adjusted)"
  416. ) %>%
  417. kable_styling(full_width = FALSE)
  418. ```
  419. ```{r Age on Stroop}
  420. # Fit model
  421. age_stroop <- lm(stroop_effect ~ age + education + project, data = merged_data)
  423. # Model summary
  424. ci_age <- summary(age_stroop)
  426. # Extract age coefficient
  427. coef_age <- ci_age$coefficients["age", ]
  428. print(ci_age)
  430. # Extract 95% CI for age
  431. ci_full <- confint(age_stroop)
  432. ci_age <- ci_full[match("age", rownames(ci_full)), ]
  434. # Raw & FDR-adjusted p-value
  435. p_raw <- coef_age["Pr(>|t|)"]
  436. p_adj <- p.adjust(p_raw, method = "fdr")
  438. # View full model summary
  439. print(ci_age)
  441. ```
  442. ##LM Analyses within and between rsfc measures
  443. Note that the manuscript focuses only on segregation as this is a measures of both within and between. Within and Between analyses and results however are reported in the supplemental materials.
  445. ```{r Mean center predictor variables}
  446. #First mean center all preditor variables
  447. merged_data <- merged_data %>%
  448. mutate(
  449. age_c = scale(age, scale = FALSE),
  450. stroop_effect_c = scale(stroop_effect, scale = FALSE),
  451. Lang_LH_within_c = scale(Lang_LH_within, scale = FALSE),
  452. MD_within_c = scale(MD_within, scale = FALSE),
  453. DMN_within_c = scale(DMN_within, scale = FALSE),
  454. Lang_to_DMN_c = scale(Lang_to_DMN, scale = FALSE),
  455. Lang_to_MD_c = scale(Lang_to_MD, scale = FALSE),
  456. DMN_to_MD_c = scale(DMN_to_MD, scale = FALSE),
  457. Lang_LH_segregation_c = scale(Lang_LH_segregation, scale = FALSE),
  458. MD_segregation_c = scale(MD_segregation, scale = FALSE),
  459. DMN_segregation_c = scale(DMN_segregation, scale = FALSE)
  460. )
  461. ```
  462. ```{r Within DMN models}
  464. #Total disfluencies
  465. model_total_disfluencies_DMN <- lm(
  466. perc_total_disfluencies ~ age_c + stroop_effect_c + DMN_within_c +
  467. age_c:stroop_effect_c + age_c:DMN_within_c + education + project,
  468. data = merged_data %>% filter(!is.na(perc_total_disfluencies)),
  469. na.action = na.exclude
  470. )
  472. #Total repetitions
  473. model_total_repetitions_DMN <- lm(
  474. perc_total_repetitions ~ age_c + stroop_effect_c + DMN_within_c +
  475. age_c:stroop_effect_c + age_c:DMN_within_c + education + project,
  476. data = merged_data %>% filter(!is.na(perc_total_repetitions)),
  477. na.action = na.exclude
  478. )
  480. #Filled pauses
  481. model_filled_pauses_DMN <- lm(
  482. perc_filled_pauses ~ age_c + stroop_effect_c + DMN_within_c +
  483. age_c:stroop_effect_c + age_c:DMN_within_c + education + project,
  484. data = merged_data %>% filter(!is.na(perc_filled_pauses)),
  485. na.action = na.exclude
  486. )
  488. #Unfilled pauses
  489. model_unfilled_pauses_DMN <- lm(
  490. perc_pauses ~ age_c + stroop_effect_c + DMN_within_c +
  491. age_c:stroop_effect_c + age_c:DMN_within_c + education + project,
  492. data = merged_data %>% filter(!is.na(perc_pauses)),
  493. na.action = na.exclude
  494. )
  496. #Prolongations
  497. model_prolongation_DMN <- lm(
  498. perc_prolongation ~ age_c + stroop_effect_c + DMN_within_c +
  499. age_c:stroop_effect_c + age_c:DMN_within_c + education + project,
  500. data = merged_data %>% filter(!is.na(perc_prolongation)),
  501. na.action = na.exclude
  502. )
  504. #Revisions
  505. model_total_revisions_DMN <- lm(
  506. perc_total_revisions ~ age_c + stroop_effect_c + DMN_within_c +
  507. age_c:stroop_effect_c + age_c:DMN_within_c + education + project,
  508. data = merged_data %>% filter(!is.na(perc_total_revisions)),
  509. na.action = na.exclude
  510. )
  512. #List all models
  513. models <- list(
  514. total_disfluencies = model_total_disfluencies_DMN,
  515. total_repetitions = model_total_repetitions_DMN,
  516. filled_pauses = model_filled_pauses_DMN,
  517. unfilled_pauses = model_unfilled_pauses_DMN,
  518. prolongation = model_prolongation_DMN,
  519. total_revisions = model_total_revisions_DMN
  520. )
  522. #Print summary() and confint() for each model
  523. for(name in names(models)) {
  524. cat("\n\n### Model for", name, "\n")
  525. print(summary(models[[name]]))
  526. cat("\n95% CI for coefficients:\n")
  527. print(confint(models[[name]]))
  528. }
  530. # extract raw p-values for DMN_within_c
  531. p_raw <- sapply(models, function(m) {
  532. summary(m)$coefficients["DMN_within_c", "Pr(>|t|)"]
  533. })
  535. # apply fdr correction across 6 tests --> main effects only
  536. # (the `n=` argument forces it to use 6 even if p_raw is shorter)
  537. p_adj <- p.adjust(p_raw, method = "fdr", n = 6)
  539. cat("\n\nfdr-adjusted p-values:\n")
  540. print(p_adj)
  542. # Extract raw p-values for interaction term age_c:DMN_within_c
  543. p_raw_interactions <- sapply(models, function(m) {
  544. summary(m)$coefficients["age_c:DMN_within_c", "Pr(>|t|)"]
  545. })
  547. # Now apply FDR correction across those 6 interaction tests
  548. p_adj_interactions <- p.adjust(p_raw_interactions, method = "fdr", n = 6)
  550. cat("\n\nFDR-adjusted p-values for age_c:DMN_within_c interaction:\n")
  551. print(p_adj_interactions)
  554. ```
  555. ```{r total disfluencies interactions}
  556. library(interactions)
  558. interact_plot(
  559. model_total_disfluencies_DMN,
  560. pred = DMN_within_c,
  561. modx = age_c,
  562. interval = TRUE,
  563. x.label = "Within-network DMN FC (centered)",
  564. y.label = "Total Disfluencies (%)",
  565. modx.labels = c("Younger", "Average", "Older")
  566. )
  568. sim_slopes(
  569. model_total_disfluencies_DMN,
  570. pred = DMN_within_c,
  571. modx = age_c,
  572. jnplot = TRUE
  573. )
  574. ```
  575. ```{r repetitions}
  576. interact_plot(
  577. model_total_repetitions_DMN,
  578. pred = DMN_within_c,
  579. modx = age_c,
  580. interval = TRUE,
  581. x.label = "Within-network DMN FC (centered)",
  582. y.label = "Repetitions (%)",
  583. modx.labels = c("Younger", "Average", "Older")
  584. )
  586. sim_slopes(
  587. model_total_repetitions_DMN,
  588. pred = DMN_within_c,
  589. modx = age_c,
  590. jnplot = TRUE
  591. )
  593. ```
  594. ```{r filled pauses}
  595. interact_plot(
  596. model_filled_pauses_DMN,
  597. pred = DMN_within_c,
  598. modx = age_c,
  599. interval = TRUE,
  600. x.label = "Within-network DMN FC (centered)",
  601. y.label = "Filled Pauses (%)",
  602. modx.labels = c("Younger", "Average", "Older")
  603. )
  605. sim_slopes(
  606. model_filled_pauses_DMN,
  607. pred = DMN_within_c,
  608. modx = age_c,
  609. jnplot = TRUE
  610. )
  611. ```
  612. ```{r Wihin MD models}
  613. # Within MD --> disfluencies
  615. # Total disfluencies
  616. model_total_disfluencies_MD <- lm(
  617. perc_total_disfluencies ~ age_c + stroop_effect_c + MD_within_c +
  618. age_c:stroop_effect_c + age_c:MD_within_c + education + project,
  619. data = merged_data %>% filter(!is.na(perc_total_disfluencies)),
  620. na.action = na.exclude
  621. )
  623. # Total repetitions
  624. model_total_repetitions_MD <- lm(
  625. perc_total_repetitions ~ age_c + stroop_effect_c + MD_within_c +
  626. age_c:stroop_effect_c + age_c:MD_within_c + education + project,
  627. data = merged_data %>% filter(!is.na(perc_total_repetitions)),
  628. na.action = na.exclude
  629. )
  631. # Filled pauses
  632. model_filled_pauses_MD <- lm(
  633. perc_filled_pauses ~ age_c + stroop_effect_c + MD_within_c +
  634. age_c:stroop_effect_c + age_c:MD_within_c + education + project,
  635. data = merged_data %>% filter(!is.na(perc_filled_pauses)),
  636. na.action = na.exclude
  637. )
  639. # Unfilled pauses
  640. model_unfilled_pauses_MD <- lm(
  641. perc_pauses ~ age_c + stroop_effect_c + MD_within_c +
  642. age_c:stroop_effect_c + age_c:MD_within_c + education + project,
  643. data = merged_data %>% filter(!is.na(perc_pauses)),
  644. na.action = na.exclude
  645. )
  647. # Prolongations
  648. model_prolongation_MD <- lm(
  649. perc_prolongation ~ age_c + stroop_effect_c + MD_within_c +
  650. age_c:stroop_effect_c + age_c:MD_within_c + education + project,
  651. data = merged_data %>% filter(!is.na(perc_prolongation)),
  652. na.action = na.exclude
  653. )
  655. # Revisions
  656. model_total_revisions_MD <- lm(
  657. perc_total_revisions ~ age_c + stroop_effect_c + MD_within_c +
  658. age_c:stroop_effect_c + age_c:MD_within_c + education + project,
  659. data = merged_data %>% filter(!is.na(perc_total_revisions)),
  660. na.action = na.exclude
  661. )
  663. # List all MD models
  664. models_MD <- list(
  665. total_disfluencies = model_total_disfluencies_MD,
  666. total_repetitions = model_total_repetitions_MD,
  667. filled_pauses = model_filled_pauses_MD,
  668. unfilled_pauses = model_unfilled_pauses_MD,
  669. prolongation = model_prolongation_MD,
  670. total_revisions = model_total_revisions_MD
  671. )
  673. # Print summary() and 95% CI for each model
  674. for (name in names(models_MD)) {
  675. cat("\n\n### Model for", name, "\n")
  676. print(summary(models_MD[[name]]))
  677. cat("\n95% CI for coefficients:\n")
  678. print(confint(models_MD[[name]]))
  679. }
  681. #Extract raw p-values for the MD_within_c term
  682. p_raw_MD <- sapply(models_MD, function(m) {
  683. summary(m)$coefficients["MD_within_c", "Pr(>|t|)"]
  684. })
  686. #Apply fdr correction across 6 tests
  687. p_adj_MD <- p.adjust(p_raw_MD, method = "fdr", n = 6)
  689. cat("\n\nfdr-adjusted p-values for MD_within_c across models:\n")
  690. print(p_adj_MD)
  691. ```
  692. ```{r Within Lang}
  694. # Total disfluencies
  695. model_total_disfluencies_LANG <- lm(
  696. perc_total_disfluencies ~ age_c + stroop_effect_c + Lang_LH_within_c +
  697. age_c:stroop_effect_c + age_c:Lang_LH_within_c + education + project,
  698. data = merged_data %>% filter(!is.na(perc_total_disfluencies)),
  699. na.action = na.exclude
  700. )
  702. # Total repetitions
  703. model_total_repetitions_LANG <- lm(
  704. perc_total_repetitions ~ age_c + stroop_effect_c + Lang_LH_within_c +
  705. age_c:stroop_effect_c + age_c:Lang_LH_within_c + education + project,
  706. data = merged_data %>% filter(!is.na(perc_total_repetitions)),
  707. na.action = na.exclude
  708. )
  710. # Filled pauses
  711. model_filled_pauses_LANG <- lm(
  712. perc_filled_pauses ~ age_c + stroop_effect_c + Lang_LH_within_c +
  713. age_c:stroop_effect_c + age_c:Lang_LH_within_c + education + project,
  714. data = merged_data %>% filter(!is.na(perc_filled_pauses)),
  715. na.action = na.exclude
  716. )
  718. # Unfilled pauses
  719. model_unfilled_pauses_LANG <- lm(
  720. perc_pauses ~ age_c + stroop_effect_c + Lang_LH_within_c +
  721. age_c:stroop_effect_c + age_c:Lang_LH_within_c + education + project,
  722. data = merged_data %>% filter(!is.na(perc_pauses)),
  723. na.action = na.exclude
  724. )
  726. # Prolongations
  727. model_prolongation_LANG <- lm(
  728. perc_prolongation ~ age_c + stroop_effect_c + Lang_LH_within_c +
  729. age_c:stroop_effect_c + age_c:Lang_LH_within_c + education + project,
  730. data = merged_data %>% filter(!is.na(perc_prolongation)),
  731. na.action = na.exclude
  732. )
  734. # Revisions
  735. model_total_revisions_LANG <- lm(
  736. perc_total_revisions ~ age_c + stroop_effect_c + Lang_LH_within_c +
  737. age_c:stroop_effect_c + age_c:Lang_LH_within_c + education + project,
  738. data = merged_data %>% filter(!is.na(perc_total_revisions)),
  739. na.action = na.exclude
  740. )
  742. # List all LANGUAGE models
  743. models_LANG <- list(
  744. total_disfluencies = model_total_disfluencies_LANG,
  745. total_repetitions = model_total_repetitions_LANG,
  746. filled_pauses = model_filled_pauses_LANG,
  747. unfilled_pauses = model_unfilled_pauses_LANG,
  748. prolongation = model_prolongation_LANG,
  749. total_revisions = model_total_revisions_LANG
  750. )
  752. # Print summary() and 95% CI for each model
  753. for (name in names(models_LANG)) {
  754. cat("\n\n### Model for", name, "\n")
  755. print(summary(models_LANG[[name]]))
  756. cat("\n95% CI for coefficients:\n")
  757. print(confint(models_LANG[[name]]))
  758. }
  760. #Extract raw p-values for the Lang_LH_within_c term
  761. p_raw_LANG <- sapply(models_LANG, function(m) {
  762. summary(m)$coefficients["Lang_LH_within_c", "Pr(>|t|)"]
  763. })
  765. #Apply fdr correction across 6 tests
  766. p_adj_LANG <- p.adjust(p_raw_LANG, method = "fdr", n = 6)
  768. cat("\n\nfdr-adjusted p-values for Lang_LH_within_c across models:\n")
  769. print(p_adj_LANG)
  770. ```
  771. ```{r Between Lang to DMN }
  772. # Between Language to DMN
  773. # Total disfluencies
  774. model_total_disfluencies_LangToDMN <- lm(
  775. perc_total_disfluencies ~ age_c + stroop_effect_c + Lang_to_DMN_c +
  776. age_c:stroop_effect_c + age_c:Lang_to_DMN_c + education + project,
  777. data = merged_data %>% filter(!is.na(perc_total_disfluencies)),
  778. na.action = na.exclude
  779. )
  781. # Total repetitions
  782. model_total_repetitions_LangToDMN <- lm(
  783. perc_total_repetitions ~ age_c + stroop_effect_c + Lang_to_DMN_c +
  784. age_c:stroop_effect_c + age_c:Lang_to_DMN_c + education + project,
  785. data = merged_data %>% filter(!is.na(perc_total_repetitions)),
  786. na.action = na.exclude
  787. )
  789. # Filled pauses
  790. model_filled_pauses_LangToDMN <- lm(
  791. perc_filled_pauses ~ age_c + stroop_effect_c + Lang_to_DMN_c +
  792. age_c:stroop_effect_c + age_c:Lang_to_DMN_c + education + project,
  793. data = merged_data %>% filter(!is.na(perc_filled_pauses)),
  794. na.action = na.exclude
  795. )
  797. # Unfilled pauses
  798. model_unfilled_pauses_LangToDMN <- lm(
  799. perc_pauses ~ age_c + stroop_effect_c + Lang_to_DMN_c +
  800. age_c:stroop_effect_c + age_c:Lang_to_DMN_c + education + project,
  801. data = merged_data %>% filter(!is.na(perc_pauses)),
  802. na.action = na.exclude
  803. )
  805. # Prolongations
  806. model_prolongation_LangToDMN <- lm(
  807. perc_prolongation ~ age_c + stroop_effect_c + Lang_to_DMN_c +
  808. age_c:stroop_effect_c + age_c:Lang_to_DMN_c + education + project,
  809. data = merged_data %>% filter(!is.na(perc_prolongation)),
  810. na.action = na.exclude
  811. )
  813. # Revisions
  814. model_total_revisions_LangToDMN <- lm(
  815. perc_total_revisions ~ age_c + stroop_effect_c + Lang_to_DMN_c +
  816. age_c:stroop_effect_c + age_c:Lang_to_DMN_c + education + project,
  817. data = merged_data %>% filter(!is.na(perc_total_revisions)),
  818. na.action = na.exclude
  819. )
  821. # List all Lang -->DMN models
  822. models_LangToDMN <- list(
  823. total_disfluencies = model_total_disfluencies_LangToDMN,
  824. total_repetitions = model_total_repetitions_LangToDMN,
  825. filled_pauses = model_filled_pauses_LangToDMN,
  826. unfilled_pauses = model_unfilled_pauses_LangToDMN,
  827. prolongation = model_prolongation_LangToDMN,
  828. total_revisions = model_total_revisions_LangToDMN
  829. )
  831. # Print summary() and 95% CI for each model
  832. for (name in names(models_LangToDMN)) {
  833. cat("\n\n### Model for", name, "\n")
  834. print(summary(models_LangToDMN[[name]]))
  835. cat("\n95% CI for coefficients:\n")
  836. print(confint(models_LangToDMN[[name]]))
  837. }
  839. # get raw p-values
  840. p_raw_main <- sapply(models_LangToDMN, function(m) {
  841. summary(m)$coefficients["Lang_to_DMN_c", "Pr(>|t|)"]
  842. })
  844. p_raw_interaction <- sapply(models_LangToDMN, function(m) {
  845. summary(m)$coefficients["age_c:Lang_to_DMN_c", "Pr(>|t|)"]
  846. })
  848. # Apply FDR correction
  849. p_adj_main <- p.adjust(p_raw_main, method = "fdr", n = length(p_raw_main))
  850. p_adj_interaction <- p.adjust(p_raw_interaction, method = "fdr", n = length(p_raw_interaction))
  852. # Print results
  853. cat("\n\nFDR-adjusted p-values for Lang_to_DMN_c (main effect):\n")
  854. print(p_adj_main)
  856. cat("\n\nFDR-adjusted p-values for age_c:Lang_to_DMN_c (interaction):\n")
  857. print(p_adj_interaction)
  858. ```
  859. ```{r test interactions}
  860. library(interactions)
  862. # Plot the interaction: Lang–DMN FC × Age predicting Filled Pauses
  863. interact_plot(
  864. model_total_disfluencies_LangToDMN,
  865. pred = Lang_to_DMN_c,
  866. modx = age_c,
  867. interval = TRUE,
  868. x.label = "Lang–DMN FC (centered)",
  869. y.label = "Total disfluencies (%)",
  870. modx.labels = c("Younger", "Average", "Older")
  871. )
  873. # Simple slopes and Johnson-Neyman test
  874. sim_slopes(
  875. model_total_disfluencies_LangToDMN,
  876. pred = Lang_to_DMN_c,
  877. modx = age_c,
  878. jnplot = TRUE
  879. )
  881. ```
  882. ```{r plot interaction}
  883. # Plot the interaction: Lang–DMN FC × Age predicting Filled Pauses
  884. interact_plot(
  885. model_total_repetitions_LangToDMN,
  886. pred = Lang_to_DMN_c,
  887. modx = age_c,
  888. interval = TRUE,
  889. x.label = "Lang–DMN FC (centered)",
  890. y.label = "Total disfluencies (%)",
  891. modx.labels = c("Younger", "Average", "Older")
  892. )
  894. # Simple slopes and Johnson-Neyman test
  895. sim_slopes(
  896. model_total_repetitions_LangToDMN,
  897. pred = Lang_to_DMN_c,
  898. modx = age_c,
  899. jnplot = TRUE
  900. )
  901. ```
  902. ```{r Between Lang to MD}
  903. # Total disfluencies
  904. model_total_disfluencies_LangToMD <- lm(
  905. perc_total_disfluencies ~ age_c + stroop_effect_c + Lang_to_MD_c +
  906. age_c:stroop_effect_c + age_c:Lang_to_MD_c + education + project,
  907. data = merged_data %>% filter(!is.na(perc_total_disfluencies)),
  908. na.action = na.exclude
  909. )
  911. # Total repetitions
  912. model_total_repetitions_LangToMD <- lm(
  913. perc_total_repetitions ~ age_c + stroop_effect_c + Lang_to_MD_c +
  914. age_c:stroop_effect_c + age_c:Lang_to_MD_c + education + project,
  915. data = merged_data %>% filter(!is.na(perc_total_repetitions)),
  916. na.action = na.exclude
  917. )
  919. # Filled pauses
  920. model_filled_pauses_LangToMD <- lm(
  921. perc_filled_pauses ~ age_c + stroop_effect_c + Lang_to_MD_c +
  922. age_c:stroop_effect_c + age_c:Lang_to_MD_c + education + project,
  923. data = merged_data %>% filter(!is.na(perc_filled_pauses)),
  924. na.action = na.exclude
  925. )
  927. # Unfilled pauses
  928. model_unfilled_pauses_LangToMD <- lm(
  929. perc_pauses ~ age_c + stroop_effect_c + Lang_to_MD_c +
  930. age_c:stroop_effect_c + age_c:Lang_to_MD_c + education + project,
  931. data = merged_data %>% filter(!is.na(perc_pauses)),
  932. na.action = na.exclude
  933. )
  935. # Prolongations
  936. model_prolongation_LangToMD <- lm(
  937. perc_prolongation ~ age_c + stroop_effect_c + Lang_to_MD_c +
  938. age_c:stroop_effect_c + age_c:Lang_to_MD_c + education + project,
  939. data = merged_data %>% filter(!is.na(perc_prolongation)),
  940. na.action = na.exclude
  941. )
  943. # Revisions
  944. model_total_revisions_LangToMD <- lm(
  945. perc_total_revisions ~ age_c + stroop_effect_c + Lang_to_MD_c +
  946. age_c:stroop_effect_c + age_c:Lang_to_MD_c + education + project,
  947. data = merged_data %>% filter(!is.na(perc_total_revisions)),
  948. na.action = na.exclude
  949. )
  951. # List all Lang→MD models
  952. models_LangToMD <- list(
  953. total_disfluencies = model_total_disfluencies_LangToMD,
  954. total_repetitions = model_total_repetitions_LangToMD,
  955. filled_pauses = model_filled_pauses_LangToMD,
  956. unfilled_pauses = model_unfilled_pauses_LangToMD,
  957. prolongation = model_prolongation_LangToMD,
  958. total_revisions = model_total_revisions_LangToMD
  959. )
  961. # Print summary() and 95% CI for each model
  962. for (name in names(models_LangToMD)) {
  963. cat("\n\n### Model for", name, "\n")
  964. print(summary(models_LangToMD[[name]]))
  965. cat("\n95% CI for coefficients:\n")
  966. print(confint(models_LangToMD[[name]]))
  967. }
  969. # FDR Correction
  970. #Extract raw p-values for Lang_to_MD_c
  971. p_raw_LangToMD <- sapply(models_LangToMD, function(m) {
  972. summary(m)$coefficients["Lang_to_MD_c", "Pr(>|t|)"]
  973. })
  975. #Apply fdr correction across 6 tests
  976. p_adj_LangToMD <- p.adjust(p_raw_LangToMD, method = "fdr", n = 6)
  978. cat("\n\nfdr-adjusted p-values for Lang_to_MD_c across models:\n")
  979. print(p_adj_LangToMD)
  980. ```
  981. ```{r Between DMN to MD}
  982. #Ran this model but since there were no a priori hypotheses, this is not included in the manuscript results
  983. # Total disfluencies
  984. model_total_disfluencies_DMNtoMD <- lm(
  985. perc_total_disfluencies ~ age_c + stroop_effect_c + DMN_to_MD_c +
  986. age_c:stroop_effect_c + age_c:DMN_to_MD_c + education + project,
  987. data = merged_data %>% filter(!is.na(perc_total_disfluencies)),
  988. na.action = na.exclude
  989. )
  991. # Total repetitions
  992. model_total_repetitions_DMNtoMD <- lm(
  993. perc_total_repetitions ~ age_c + stroop_effect_c + DMN_to_MD_c +
  994. age_c:stroop_effect_c + age_c:DMN_to_MD_c + education + project,
  995. data = merged_data %>% filter(!is.na(perc_total_repetitions)),
  996. na.action = na.exclude
  997. )
  999. # Filled pauses
  1000. model_filled_pauses_DMNtoMD <- lm(
  1001. perc_filled_pauses ~ age_c + stroop_effect_c + DMN_to_MD_c +
  1002. age_c:stroop_effect_c + age_c:DMN_to_MD_c + education + project,
  1003. data = merged_data %>% filter(!is.na(perc_filled_pauses)),
  1004. na.action = na.exclude
  1005. )
  1007. # Unfilled pauses
  1008. model_unfilled_pauses_DMNtoMD <- lm(
  1009. perc_pauses ~ age_c + stroop_effect_c + DMN_to_MD_c +
  1010. age_c:stroop_effect_c + age_c:DMN_to_MD_c + education + project,
  1011. data = merged_data %>% filter(!is.na(perc_pauses)),
  1012. na.action = na.exclude
  1013. )
  1015. # Prolongations
  1016. model_prolongation_DMNtoMD <- lm(
  1017. perc_prolongation ~ age_c + stroop_effect_c + DMN_to_MD_c +
  1018. age_c:stroop_effect_c + age_c:DMN_to_MD_c + education + project,
  1019. data = merged_data %>% filter(!is.na(perc_prolongation)),
  1020. na.action = na.exclude
  1021. )
  1023. # Revisions
  1024. model_total_revisions_DMNtoMD <- lm(
  1025. perc_total_revisions ~ age_c + stroop_effect_c + DMN_to_MD_c +
  1026. age_c:stroop_effect_c + age_c:DMN_to_MD_c + education + project,
  1027. data = merged_data %>% filter(!is.na(perc_total_revisions)),
  1028. na.action = na.exclude
  1029. )
  1031. # List all DMN to MD models
  1032. models_DMNtoMD <- list(
  1033. total_disfluencies = model_total_disfluencies_DMNtoMD,
  1034. total_repetitions = model_total_repetitions_DMNtoMD,
  1035. filled_pauses = model_filled_pauses_DMNtoMD,
  1036. unfilled_pauses = model_unfilled_pauses_DMNtoMD,
  1037. prolongation = model_prolongation_DMNtoMD,
  1038. total_revisions = model_total_revisions_DMNtoMD
  1039. )
  1041. # Print summary() and 95% CI for each model
  1042. for (name in names(models_DMNtoMD)) {
  1043. cat("\n\n### Model for", name, "\n")
  1044. print(summary(models_DMNtoMD[[name]]))
  1045. cat("\n95% CI for coefficients:\n")
  1046. print(confint(models_DMNtoMD[[name]]))
  1047. }
  1049. #FDR correction
  1050. #Extract raw p-values for the DMN_to_MD_c term
  1051. p_raw_DMNtoMD <- sapply(models_DMNtoMD, function(m) {
  1052. summary(m)$coefficients["DMN_to_MD_c", "Pr(>|t|)"]
  1053. })
  1055. #Apply fdr correction across 6 tests
  1056. p_adj_DMNtoMD <- p.adjust(p_raw_DMNtoMD, method = "fdr", n = 6)
  1058. cat("\n\nfdr-adjusted p-values for DMN_to_MD_c across models:\n")
  1059. print(p_adj_DMNtoMD)
  1060. ```
  1061. #LM Analyses Segregation rsfc measure
  1062. Here is the main analyses reported in the manuscript i.e., the Segregation measure
  1063. ```{r Segregation DMN}
  1064. # Total disfluencies
  1065. model_total_disfluencies_DMNseg <- lm(
  1066. perc_total_disfluencies ~ age_c + stroop_effect_c + DMN_segregation_c +
  1067. age_c:stroop_effect_c + age_c:DMN_segregation_c + education + project,
  1068. data = merged_data %>% filter(!is.na(perc_total_disfluencies)),
  1069. na.action = na.exclude
  1070. )
  1072. # Total repetitions
  1073. model_total_repetitions_DMNseg <- lm(
  1074. perc_total_repetitions ~ age_c + stroop_effect_c + DMN_segregation_c +
  1075. age_c:stroop_effect_c + age_c:DMN_segregation_c + education + project,
  1076. data = merged_data %>% filter(!is.na(perc_total_repetitions)),
  1077. na.action = na.exclude
  1078. )
  1080. # Filled pauses
  1081. model_filled_pauses_DMNseg <- lm(
  1082. perc_filled_pauses ~ age_c + stroop_effect_c + DMN_segregation_c +
  1083. age_c:stroop_effect_c + age_c:DMN_segregation_c + education + project,
  1084. data = merged_data %>% filter(!is.na(perc_filled_pauses)),
  1085. na.action = na.exclude
  1086. )
  1088. # Unfilled pauses
  1089. model_unfilled_pauses_DMNseg <- lm(
  1090. perc_pauses ~ age_c + stroop_effect_c + DMN_segregation_c +
  1091. age_c:stroop_effect_c + age_c:DMN_segregation_c + education + project,
  1092. data = merged_data %>% filter(!is.na(perc_pauses)),
  1093. na.action = na.exclude
  1094. )
  1096. # Prolongations
  1097. model_prolongation_DMNseg <- lm(
  1098. perc_prolongation ~ age_c + stroop_effect_c + DMN_segregation_c +
  1099. age_c:stroop_effect_c + age_c:DMN_segregation_c + education + project,
  1100. data = merged_data %>% filter(!is.na(perc_prolongation)),
  1101. na.action = na.exclude
  1102. )
  1104. # Revisions
  1105. model_total_revisions_DMNseg <- lm(
  1106. perc_total_revisions ~ age_c + stroop_effect_c + DMN_segregation_c +
  1107. age_c:stroop_effect_c + age_c:DMN_segregation_c + education + project,
  1108. data = merged_data %>% filter(!is.na(perc_total_revisions)),
  1109. na.action = na.exclude
  1110. )
  1112. # List all DMN segregation models
  1113. models_DMNseg <- list(
  1114. total_disfluencies = model_total_disfluencies_DMNseg,
  1115. total_repetitions = model_total_repetitions_DMNseg,
  1116. filled_pauses = model_filled_pauses_DMNseg,
  1117. unfilled_pauses = model_unfilled_pauses_DMNseg,
  1118. prolongation = model_prolongation_DMNseg,
  1119. total_revisions = model_total_revisions_DMNseg
  1120. )
  1122. # Print summary() and 95% CI for each model
  1123. for (name in names(models_DMNseg)) {
  1124. cat("\n\n### Model for", name, "\n")
  1125. print(summary(models_DMNseg[[name]]))
  1126. cat("\n95% CI for coefficients:\n")
  1127. print(confint(models_DMNseg[[name]]))
  1128. }
  1129. ```
  1130. ```{r Test Interactions Age x DMNseg}
  1131. # Model
  1132. model_dmn_total <- lm(
  1133. perc_total_disfluencies ~ age_c + DMN_segregation_c +
  1134. stroop_effect_c + age_c:DMN_segregation_c +
  1135. age_c:stroop_effect_c + education + project,
  1136. data = merged_data
  1137. )
  1139. # Plot interaction
  1140. interact_plot(
  1141. model_dmn_total,
  1142. pred = DMN_segregation_c,
  1143. modx = age_c,
  1144. interval = TRUE,
  1145. x.label = "DMN Segregation (centered)",
  1146. y.label = "Total Disfluencies (%)",
  1147. modx.labels = c("Younger", "Average", "Older")
  1148. )
  1150. # Simple slopes and Johnson-Neyman
  1151. sim_slopes(
  1152. model_dmn_total,
  1153. pred = DMN_segregation_c,
  1154. modx = age_c,
  1155. jnplot = TRUE
  1156. )
  1157. ```
  1158. ```{r Test Interactions Age x DMNseg}
  1159. model_dmn_reps <- lm(
  1160. perc_total_repetitions ~ age_c + DMN_segregation_c +
  1161. stroop_effect_c + age_c:DMN_segregation_c +
  1162. age_c:stroop_effect_c + education + project,
  1163. data = merged_data
  1164. )
  1166. interact_plot(
  1167. model_dmn_reps,
  1168. pred = DMN_segregation_c,
  1169. modx = age_c,
  1170. interval = TRUE,
  1171. x.label = "DMN Segregation (centered)",
  1172. y.label = "Repetitions (%)",
  1173. modx.labels = c("Younger", "Average", "Older")
  1174. )
  1176. sim_slopes(
  1177. model_dmn_reps,
  1178. pred = DMN_segregation_c,
  1179. modx = age_c,
  1180. jnplot = TRUE
  1181. )
  1183. ```
  1184. ```{r Segregation MD}
  1185. # Total disfluencies
  1186. model_total_disfluencies_MDseg <- lm(
  1187. perc_total_disfluencies ~ age_c + stroop_effect_c + MD_segregation_c +
  1188. age_c:stroop_effect_c + age_c:MD_segregation_c + education + project,
  1189. data = merged_data %>% filter(!is.na(perc_total_disfluencies)),
  1190. na.action = na.exclude
  1191. )
  1193. # Total repetitions
  1194. model_total_repetitions_MDseg <- lm(
  1195. perc_total_repetitions ~ age_c + stroop_effect_c + MD_segregation_c +
  1196. age_c:stroop_effect_c + age_c:MD_segregation_c + education + project,
  1197. data = merged_data %>% filter(!is.na(perc_total_repetitions)),
  1198. na.action = na.exclude
  1199. )
  1201. # Filled pauses
  1202. model_filled_pauses_MDseg <- lm(
  1203. perc_filled_pauses ~ age_c + stroop_effect_c + MD_segregation_c +
  1204. age_c:stroop_effect_c + age_c:MD_segregation_c + education + project,
  1205. data = merged_data %>% filter(!is.na(perc_filled_pauses)),
  1206. na.action = na.exclude
  1207. )
  1209. # Unfilled pauses
  1210. model_unfilled_pauses_MDseg <- lm(
  1211. perc_pauses ~ age_c + stroop_effect_c + MD_segregation_c +
  1212. age_c:stroop_effect_c + age_c:MD_segregation_c + education + project,
  1213. data = merged_data %>% filter(!is.na(perc_pauses)),
  1214. na.action = na.exclude
  1215. )
  1217. # Prolongations
  1218. model_prolongation_MDseg <- lm(
  1219. perc_prolongation ~ age_c + stroop_effect_c + MD_segregation_c +
  1220. age_c:stroop_effect_c + age_c:MD_segregation_c + education + project,
  1221. data = merged_data %>% filter(!is.na(perc_prolongation)),
  1222. na.action = na.exclude
  1223. )
  1225. # Revisions
  1226. model_total_revisions_MDseg <- lm(
  1227. perc_total_revisions ~ age_c + stroop_effect_c + MD_segregation_c +
  1228. age_c:stroop_effect_c + age_c:MD_segregation_c + education + project,
  1229. data = merged_data %>% filter(!is.na(perc_total_revisions)),
  1230. na.action = na.exclude
  1231. )
  1233. # List all MD segregation models
  1234. models_MDseg <- list(
  1235. total_disfluencies = model_total_disfluencies_MDseg,
  1236. total_repetitions = model_total_repetitions_MDseg,
  1237. filled_pauses = model_filled_pauses_MDseg,
  1238. unfilled_pauses = model_unfilled_pauses_MDseg,
  1239. prolongation = model_prolongation_MDseg,
  1240. total_revisions = model_total_revisions_MDseg
  1241. )
  1243. # Print summary() and 95% CI for each model
  1244. for (name in names(models_MDseg)) {
  1245. cat("\n\n### Model for", name, "\n")
  1246. print(summary(models_MDseg[[name]]))
  1247. cat("\n95% CI for coefficients:\n")
  1248. print(confint(models_MDseg[[name]]))
  1249. }
  1251. ```
  1252. ```{r Segregation Lang}
  1253. # Total disfluencies
  1254. model_total_disfluencies_LANGseg <- lm(
  1255. perc_total_disfluencies ~ age_c + stroop_effect_c + Lang_LH_segregation_c +
  1256. age_c:stroop_effect_c + age_c:Lang_LH_segregation_c + education + project,
  1257. data = merged_data %>% filter(!is.na(perc_total_disfluencies)),
  1258. na.action = na.exclude
  1259. )
  1261. # Total repetitions
  1262. model_total_repetitions_LANGseg <- lm(
  1263. perc_total_repetitions ~ age_c + stroop_effect_c + Lang_LH_segregation_c +
  1264. age_c:stroop_effect_c + age_c:Lang_LH_segregation_c + education + project,
  1265. data = merged_data %>% filter(!is.na(perc_total_repetitions)),
  1266. na.action = na.exclude
  1267. )
  1269. # Filled pauses
  1270. model_filled_pauses_LANGseg <- lm(
  1271. perc_filled_pauses ~ age_c + stroop_effect_c + Lang_LH_segregation_c +
  1272. age_c:stroop_effect_c + age_c:Lang_LH_segregation_c + education + project,
  1273. data = merged_data %>% filter(!is.na(perc_filled_pauses)),
  1274. na.action = na.exclude
  1275. )
  1277. # Unfilled pauses
  1278. model_unfilled_pauses_LANGseg <- lm(
  1279. perc_pauses ~ age_c + stroop_effect_c + Lang_LH_segregation_c +
  1280. age_c:stroop_effect_c + age_c:Lang_LH_segregation_c + education + project,
  1281. data = merged_data %>% filter(!is.na(perc_pauses)),
  1282. na.action = na.exclude
  1283. )
  1285. # Prolongations
  1286. model_prolongation_LANGseg <- lm(
  1287. perc_prolongation ~ age_c + stroop_effect_c + Lang_LH_segregation_c +
  1288. age_c:stroop_effect_c + age_c:Lang_LH_segregation_c + education + project,
  1289. data = merged_data %>% filter(!is.na(perc_prolongation)),
  1290. na.action = na.exclude
  1291. )
  1293. # Revisions
  1294. model_total_revisions_LANGseg <- lm(
  1295. perc_total_revisions ~ age_c + stroop_effect_c + Lang_LH_segregation_c +
  1296. age_c:stroop_effect_c + age_c:Lang_LH_segregation_c + education + project,
  1297. data = merged_data %>% filter(!is.na(perc_total_revisions)),
  1298. na.action = na.exclude
  1299. )
  1301. # List all LANGUAGE segregation models
  1302. models_LANGseg <- list(
  1303. total_disfluencies = model_total_disfluencies_LANGseg,
  1304. total_repetitions = model_total_repetitions_LANGseg,
  1305. filled_pauses = model_filled_pauses_LANGseg,
  1306. unfilled_pauses = model_unfilled_pauses_LANGseg,
  1307. prolongation = model_prolongation_LANGseg,
  1308. total_revisions = model_total_revisions_LANGseg
  1309. )
  1311. # Print summary() and 95% CI for each model
  1312. for (name in names(models_LANGseg)) {
  1313. cat("\n\n### Model for", name, "\n")
  1314. print(summary(models_LANGseg[[name]]))
  1315. cat("\n95% CI for coefficients:\n")
  1316. print(confint(models_LANGseg[[name]]))
  1317. }
  1319. ```
  1320. #Correct for Multiple Comparisons Main Effects
  1321. ```{r fdr correction}
  1323. #Extract raw p-values for each network’s 6 disfluency tests
  1324. p_DMN <- sapply(models_DMNseg, function(m) coef(summary(m))["DMN_segregation_c", "Pr(>|t|)"])
  1325. p_MD <- sapply(models_MDseg, function(m) coef(summary(m))["MD_segregation_c", "Pr(>|t|)"])
  1326. p_LANG <- sapply(models_LANGseg, function(m) coef(summary(m))["Lang_LH_segregation_c","Pr(>|t|)"])
  1328. #FDR-adjust within each network
  1329. q_DMN <- p.adjust(p_DMN, method = "fdr")
  1330. q_MD <- p.adjust(p_MD, method = "fdr")
  1331. q_LANG <- p.adjust(p_LANG, method = "fdr")
  1333. #Make a combined table
  1334. seg_by_net <- bind_rows(
  1335. tibble(
  1336. network = "DMN",
  1337. outcome = names(p_DMN),
  1338. p_raw = unname(p_DMN),
  1339. q_fdr = unname(q_DMN)
  1340. ),
  1341. tibble(
  1342. network = "MD",
  1343. outcome = names(p_MD),
  1344. p_raw = unname(p_MD),
  1345. q_fdr = unname(q_MD)
  1346. ),
  1347. tibble(
  1348. network = "LANG",
  1349. outcome = names(p_LANG),
  1350. p_raw = unname(p_LANG),
  1351. q_fdr = unname(q_LANG)
  1352. )
  1353. )
  1355. #Print table
  1356. seg_by_net %>%
  1357. arrange(network, q_fdr) %>%
  1358. kable(
  1359. digits = 3,
  1360. caption = "Network-specific Segregation → Disfluency: raw p and FDR-q"
  1361. ) %>%
  1362. kable_styling(full_width = FALSE)
  1363. ```
  1364. #Correct for Multiple Comparisons: Interactions
  1365. ```{r fdr correction interactions}
  1366. #Extract six interaction p-values per network
  1367. pint_DMN <- sapply(models_DMNseg, function(m) coef(summary(m))["age_c:DMN_segregation_c", "Pr(>|t|)"])
  1368. pint_MD <- sapply(models_MDseg, function(m) coef(summary(m))["age_c:MD_segregation_c", "Pr(>|t|)"])
  1369. pint_LANG <- sapply(models_LANGseg, function(m) coef(summary(m))["age_c:Lang_LH_segregation_c","Pr(>|t|)"])
  1371. #Apply FDR correction
  1372. q_int_DMN <- p.adjust(pint_DMN, method = "fdr")
  1373. q_int_MD <- p.adjust(pint_MD, method = "fdr")
  1374. q_int_LANG <- p.adjust(pint_LANG, method = "fdr")
  1376. #Make a table
  1377. int_by_net <- bind_rows(
  1378. tibble(network = "DMN", outcome = names(pint_DMN), p_int = pint_DMN, q_int = q_int_DMN),
  1379. tibble(network = "MD", outcome = names(pint_MD), p_int = pint_MD, q_int = q_int_MD),
  1380. tibble(network = "LANG", outcome = names(pint_LANG), p_int = pint_LANG, q_int = q_int_LANG)
  1381. )
  1383. #print
  1384. int_by_net %>%
  1385. arrange(network, q_int) %>%
  1386. kable(
  1387. digits = 3,
  1388. caption = "Network‐specific Segregation × Age Interactions: raw p and FDR‐q"
  1389. ) %>%
  1390. kable_styling(full_width = FALSE)
  1392. ```
  1395. #Start of mediation analyses
  1396. First we will see if disfluency is mediated by Stroop, then DMN segregation, then Lang-to-DMN between
  1397. ```{r load libraries}
  1398. library(mediation)
  1399. ```
  1400. ##Mediation A: Stroop mediatior for Total Disfluencies
  1401. Age --> Stroop --> Total_Disfluency
  1402. ```{r mediation model A}
  1403. #Path 1: Age --> EF (Stroop)
  1404. model_A1 <- lm(stroop_effect ~ age + education +project, data = merged_data)
  1406. #Path 2: EF (Stroop) --> Total Disfluencies (controlling for Age)
  1407. model_A2 <- lm(perc_total_disfluencies ~ stroop_effect + age + education +project, data = merged_data)
  1409. #Path 3: Age --> Total Disfluencies
  1410. model_A3 <- lm(perc_total_disfluencies ~ age + education + project, data = merged_data)
  1412. summary(model_A1)
  1413. summary(model_A2)
  1414. summary(model_A3)
  1415. ```
  1416. ##Mediation B: Stroop mediatior for Revisions
  1417. Age --> Stroop --> Revisions
  1419. Now I am going to check just revisions since this overall seems to be the disfluency type that was most sensitive
  1420. ```{r mediation model B}
  1421. #Path 1: Age --> Stroop (same)
  1422. model_B1_rev <- lm(stroop_effect ~ age + education + project, data = merged_data)
  1424. #Path 2: Stroop --> Revisions (controlling for Age)
  1425. model_B2_rev <- lm(perc_total_revisions ~ stroop_effect + age + education + project, data = merged_data)
  1427. # Path 3: Age --> Revisions
  1428. model_B3_rev <- lm(perc_total_revisions ~ age + education + project, data = merged_data)
  1430. summary(model_B1_rev)
  1431. summary(model_B2_rev)
  1432. summary(model_B3_rev)
  1434. ```
  1435. ```{r Full mediation model B}
  1436. # Filter to missing data
  1437. modelB_data <- merged_data %>%
  1438. filter(!is.na(age), !is.na(education), !is.na(stroop_effect), !is.na(perc_total_revisions))
  1440. #re-fit models
  1441. modelB_mediator <- lm(stroop_effect ~ age + education + project, data = modelB_data)
  1442. modelB_outcome <- lm(perc_total_revisions ~ stroop_effect + age + education + project, data = modelB_data)
  1444. #Mediation analysis
  1445. library(mediation)
  1446. set.seed(1234)
  1447. modelB_med <- mediate(
  1448. model.m = modelB_mediator,
  1449. model.y = modelB_outcome,
  1450. treat = "age",
  1451. mediator = "stroop_effect",
  1452. boot = TRUE,
  1453. sims = 5000
  1454. )
  1456. summary(modelB_med)
  1457. ```
  1460. ##Mediation C: Stroop mediator for Filled Pauses
  1461. Age --> Stroop --> Filled Pauses
  1462. ```{r mediation C}
  1463. #Path 1: Age --> Stroop
  1464. model_C1_fp <- lm(stroop_effect ~ age + education + project, data = merged_data)
  1466. #Path 2: Stroop --> Filled Pauses (controlling for Age)
  1467. model_C2_fp <- lm(perc_filled_pauses ~ stroop_effect + age + education + project, data = merged_data)
  1469. #Path 3: Age --> Filled Pauses
  1470. model_C3_fp <- lm(perc_filled_pauses ~ age + education + project, data = merged_data)
  1472. summary(model_C1_fp)
  1473. summary(model_C2_fp)
  1474. summary(model_C3_fp)
  1475. ```
  1476. ```{r Full Mediation model C}
  1477. # Filter to missing data
  1478. modelC_data <- merged_data %>%
  1479. filter(!is.na(age), !is.na(education), !is.na(stroop_effect), !is.na(perc_filled_pauses))
  1481. #re-fit models
  1482. modelC_mediator <- lm(stroop_effect ~ age + education + project, data = modelC_data)
  1483. modelC_outcome <- lm(perc_filled_pauses ~ stroop_effect + age + education + project, data = modelC_data)
  1485. #Mediation analysis
  1486. library(mediation)
  1487. set.seed(1234)
  1488. modelC_med <- mediate(
  1489. model.m = modelC_mediator,
  1490. model.y = modelC_outcome,
  1491. treat = "age",
  1492. mediator = "stroop_effect",
  1493. boot = TRUE,
  1494. sims = 5000
  1495. )
  1497. summary(modelC_med)
  1498. ```
  1499. ##Mediation D: Stroop mediator for Unfilled Pauses
  1500. Age --> Stroop --> Unfilled Pauses
  1501. ```{r Mediation model D}
  1502. #Path 1: Age --> Stroop
  1503. model_D1_sp <- lm(stroop_effect ~ age + education + project, data = merged_data)
  1505. #Path 2: Stroop --> Silent Pauses (controlling for Age)
  1506. model_D2_sp <- lm(perc_pauses ~ stroop_effect + age + education + project, data = merged_data)
  1508. #Path 3: Age --> Silent Pauses
  1509. model_D3_sp <- lm(perc_pauses ~ age + education + project, data = merged_data)
  1511. summary(model_D1_sp)
  1512. summary(model_D2_sp)
  1513. summary(model_D3_sp)
  1515. ```
  1516. ```{r full mediation model D}
  1517. # Filter to missing data
  1518. modelD_data <- merged_data %>%
  1519. filter(!is.na(age), !is.na(education), !is.na(stroop_effect), !is.na(perc_pauses))
  1521. #re-fit models
  1522. modelD_mediator <- lm(stroop_effect ~ age + education + project, data = modelD_data)
  1523. modelD_outcome <- lm(perc_filled_pauses ~ stroop_effect + age + education + project, data = modelD_data)
  1525. #Mediation analysis
  1526. library(mediation)
  1527. set.seed(1234)
  1528. modelD_med <- mediate(
  1529. model.m = modelD_mediator,
  1530. model.y = modelD_outcome,
  1531. treat = "age",
  1532. mediator = "stroop_effect",
  1533. boot = TRUE,
  1534. sims = 5000
  1535. )
  1537. summary(modelD_med)
  1538. ```
  1539. #Mediation Analysis RSFC Measures
  1541. Based on the LM analyses results the best candidates identified were:
  1542. Age --> DMN_Segregation --> Revisions
  1543. Age --> Lang_to_DMN --> Revisions
  1544. Age --> Lang_to-MD --> Unfilled Pauses
  1546. Do not meet mediation criteria:
  1547. - MD segregation: Age-related change, but no FC --> disfluency effect
  1548. - MD within / Lang within / Lang segregation: No FC --> disfluency link at all
  1549. - DMN-to-MD: Not predictive of disfluency
  1550. - Lang within: No effects
  1551. - MD within: No effects
  1552. - DMN segregation → total disfluencies: Only marginal effect
  1554. ##Mediation Model E: DMN segregation
  1555. ```{r Mediation model E check}
  1556. # Path 1: Age → DMN segregation
  1557. model_E1 <- lm(DMN_segregation_c ~ age + education + project, data = merged_data)
  1559. # Path 2: DMN segregation → Revisions (controlling for Age)
  1560. model_E2 <- lm(perc_total_revisions ~ DMN_segregation_c + age + education + project, data = merged_data)
  1562. # Path 3: Age → Revisions
  1563. model_E3 <- lm(perc_total_revisions ~ age + education + project, data = merged_data)
  1565. summary(model_E1)
  1566. summary(model_E2)
  1567. summary(model_E3)
  1568. ```
  1569. ```{r Full mediation model E}
  1570. # Filter to complete cases for relevant variables
  1571. med_data_E <- merged_data %>%
  1572. filter(!is.na(age), !is.na(education), !is.na(DMN_segregation_c), !is.na(perc_total_revisions))
  1574. # Make sure the mediator is treated as numeric
  1575. merged_data$DMN_segregation_c <- as.numeric(merged_data$DMN_segregation_c)
  1577. # Refit the models
  1578. modelE_mediator <- lm(DMN_segregation_c ~ age + education + project, data = merged_data)
  1579. modelE_outcome <- lm(perc_total_revisions ~ DMN_segregation_c + age + education + project, data = merged_data)
  1581. # Run mediation
  1582. library(mediation)
  1583. set.seed(1234)
  1584. modelE_med <- mediate(
  1585. model.m = modelE_mediator,
  1586. model.y = modelE_outcome,
  1587. treat = "age",
  1588. mediator = "DMN_segregation_c",
  1589. boot = TRUE,
  1590. sims = 5000
  1591. )
  1593. # Summarize results
  1594. summary(modelE_med)
  1596. ```
  1597. ##Mediation model F: Lang_to_DMN
  1599. ```{r Mediation model F}
  1600. # Path 1: Age → Lang_to_DMN
  1601. #model_F1 <- lm(Lang_to_DMN ~ age + education + project, data = merged_data)
  1603. # Path 2: Lang_to_DMN → Revisions (controlling for Age)
  1604. #model_F2 <- lm(perc_total_revisions ~ Lang_to_DMN + age + education + project, data = merged_data)
  1606. # Path 3: Age → Revisions
  1607. #model_F3 <- lm(perc_total_revisions ~ age + education + project, data = merged_data)
  1609. #summary(model_F1)
  1610. #summary(model_F2)
  1611. #summary(model_F3)
  1613. ```
  1614. ```{r Full mediation model F}
  1615. # Step 1: Subset to complete cases for relevant variables
  1616. modelF_data <- merged_data %>%
  1617. filter( !is.na(age), !is.na(education), !is.na(Lang_to_DMN), !is.na(perc_total_revisions)
  1618. )
  1620. # Step 2: Fit the mediator model (Path a)
  1621. mediator_model_F <- lm(Lang_to_DMN ~ age + education + project, data = modelF_data)
  1623. # Step 3: Fit the outcome model (Paths b and c′)
  1624. outcome_model_F <- lm(perc_total_revisions ~ Lang_to_DMN + age + education + project, data = modelF_data)
  1626. # Step 4: Run mediation analysis
  1627. set.seed(1234)
  1628. mediation_result_F <- mediate(
  1629. model.m = mediator_model_F,
  1630. model.y = outcome_model_F,
  1631. treat = "age",
  1632. mediator = "Lang_to_DMN",
  1633. boot = TRUE,
  1634. sims = 5000
  1635. )
  1637. # Step 5: View results
  1638. summary(mediation_result_F)
  1639. ```
  1640. ##Mediation model G: Lang_to_DMN
  1642. ```{r Mediation model G}
  1643. # Path 1: Age → Lang_to_MD
  1644. model_G1 <- lm(Lang_to_MD ~ age + education + project, data = merged_data)
  1646. # Path 2: Lang_to_MD → Unfilled Pauses (controlling for Age)
  1647. model_G2 <- lm(perc_pauses ~ Lang_to_MD + age + education + project, data = merged_data)
  1649. # Path 3: Age → Unfilled Pauses
  1650. model_G3 <- lm(perc_pauses ~ age + education + project, data = merged_data)
  1652. summary(model_G1)
  1653. summary(model_G2)
  1654. summary(model_G3)
  1655. ```
  1656. ##Mediation model H: Stroop on repetitions
  1657. ```{r mediation model H}
  1658. # Path a: Age --> Stroop
  1659. model_Ha <- lm(stroop_effect ~ age + education + project, data = merged_data)
  1661. # Path b: Stroop --> Repetitions
  1662. model_Hb <- lm(perc_total_repetitions ~ stroop_effect + age + education + project, data = merged_data)
  1664. # Path c: Age --> Repetitions
  1665. model_Hc <- lm(perc_total_repetitions ~ age + education + project, data = merged_data)
  1668. summary(model_Ha)
  1669. summary(model_Hb)
  1670. summary(model_Hc)
  1672. ```
  1673. ```{r full mediation model H}
  1674. # Filter to missing data
  1675. modelH_data <- merged_data %>%
  1676. filter(!is.na(age), !is.na(education), !is.na(stroop_effect), !is.na(perc_pauses))
  1678. #re-fit models
  1679. modelH_mediator <- lm(stroop_effect ~ age + education + project, data = modelH_data)
  1680. modelH_outcome <- lm(perc_filled_pauses ~ stroop_effect + age + education + project, data = modelH_data)
  1682. #Mediation analysis
  1683. library(mediation)
  1684. set.seed(1234)
  1685. modelH_med <- mediate(
  1686. model.m = modelH_mediator,
  1687. model.y = modelH_outcome,
  1688. treat = "age",
  1689. mediator = "stroop_effect",
  1690. boot = TRUE,
  1691. sims = 5000
  1692. )
  1694. summary(modelH_med)
  1695. ```
  1696. ##Mediation model I: DMN seg on repetitions
  1697. ```{r mediation model I}
  1698. # Path a: Age --> DMN segregation
  1699. model_Ia <- lm(DMN_segregation_c ~ age + education + project, data = merged_data)
  1701. # Path b: DMN segregation --> Repetitions
  1702. model_Ib <- lm(perc_total_repetitions ~ DMN_segregation_c + age + education + project, data = merged_data)
  1704. # Path c: Age --> Repetitions
  1705. model_Ic <- lm(perc_total_repetitions ~ age + education + project, data = merged_data)
  1707. # Review model summaries
  1708. summary(model_Ia)
  1709. summary(model_Ib)
  1710. summary(model_Ic)
  1712. ```
  1713. ```{r full mediation model I}
  1714. # Subset complete cases
  1715. modelI_data <- merged_data %>%
  1716. filter(!is.na(age), !is.na(education), !is.na(project),
  1717. !is.na(DMN_segregation_c), !is.na(perc_total_repetitions))
  1719. # Convert to numeric (fixes the error)
  1720. modelI_data$DMN_segregation_c <- as.numeric(modelI_data$DMN_segregation_c)
  1722. # Refit models
  1723. modelI_mediator <- lm(DMN_segregation_c ~ age + education + project, data = modelI_data)
  1724. modelI_outcome <- lm(perc_total_repetitions ~ DMN_segregation_c + age + education + project, data = modelI_data)
  1726. # Run mediation
  1727. library(mediation)
  1728. set.seed(1234)
  1729. modelI_med <- mediate(
  1730. model.m = modelI_mediator,
  1731. model.y = modelI_outcome,
  1732. treat = "age",
  1733. mediator = "DMN_segregation_c",
  1734. boot = TRUE,
  1735. sims = 5000
  1736. )
  1738. # Display results
  1739. summary(modelI_med)
  1741. ```
  1742. #Mediation model J: Lang_to_DMN on repetitions
  1744. ```{r mediation model J repetitions}
  1745. # Path a: Age --> Lang-to-DMN
  1746. model_Ja <- lm(Lang_to_DMN ~ age + education + project, data = merged_data)
  1748. # Path b: Lang-to-DMN --> Repetitions
  1749. model_Jb <- lm(perc_total_repetitions ~ Lang_to_DMN + age + education + project, data = merged_data)
  1751. # Path c: Age --> Repetitions
  1752. model_Jc <- lm(perc_total_repetitions ~ age + education + project, data = merged_data)
  1754. # Review model summaries
  1755. summary(model_Ja)
  1756. summary(model_Jb)
  1757. summary(model_Jc)
  1759. ```
  1760. ```{r full mediation model J}
  1761. # Subset complete cases
  1762. modelJ_data <- merged_data %>%
  1763. filter(!is.na(age), !is.na(education), !is.na(project),
  1764. !is.na(Lang_to_DMN), !is.na(perc_total_repetitions))
  1766. # Convert to numeric (fixes potential mediation error)
  1767. modelJ_data$Lang_to_DMN <- as.numeric(modelJ_data$Lang_to_DMN)
  1769. # Refit models
  1770. modelJ_mediator <- lm(Lang_to_DMN ~ age + education + project, data = modelJ_data)
  1771. modelJ_outcome <- lm(perc_total_repetitions ~ Lang_to_DMN + age + education + project, data = modelJ_data)
  1773. # Run mediation
  1774. library(mediation)
  1775. set.seed(1234)
  1776. modelJ_med <- mediate(
  1777. model.m = modelJ_mediator,
  1778. model.y = modelJ_outcome,
  1779. treat = "age",
  1780. mediator = "Lang_to_DMN",
  1781. boot = TRUE,
  1782. sims = 5000
  1783. )
  1785. # Display results
  1786. summary(modelJ_med)
  1787. ```
  1788. #Serial Mediation check: Stroop ~ DMN segregation + age + education
  1789. ```{r lm check}
  1790. model_stroop_DMNseg <- lm(stroop_effect ~ DMN_segregation_c + age + education +project, data = merged_data)
  1791. summary(model_stroop_DMNseg)
  1792. ```
  1793. #Serial Mediation check: Stroop ~ Lang_to_DMN + age + education
  1794. ```{r lm check}
  1795. model_stroop_LangDMN <- lm(stroop_effect ~ Lang_to_DMN + age + education + project, data = merged_data)
  1796. summary(model_stroop_LangDMN)
  1797. ```
  1798. #Serial Mediation Analysis
  1799. ```{r serial mediation}
  1800. # Model 1: Mediator model (Stroop ~ Age + RSFC + education)
  1801. model_m <- lm(stroop_effect ~ age + DMN_segregation + education + project, data = merged_data)
  1803. # Model 2: Outcome model (Disfluency ~ Age + Stroop + RSFC + education)
  1804. model_y <- lm(perc_total_disfluencies ~ age + stroop_effect + DMN_segregation + education + project, data = merged_data)
  1806. # Run mediation analysis
  1807. library(mediation)
  1808. med_out <- mediate(model.m = model_m, model.y = model_y,
  1809. treat = "age", mediator = "stroop_effect",
  1810. boot = TRUE, sims = 5000)
  1812. summary(med_out)
  1813. ```

RSFC_stats_analysis_OSF.Rmd, no license · at the source

Overview

Authors: Megan S Nakamura1, Haoyun Zhang2, Michele T Diaz1
ORCID iDs: Haoyun Zhang, Michele T Diaz
  1. The Pennsylvania State University, University Park, PA
  2. Centre for Cognitive and Brain Sciences, Department of Psychology, University of Macau, Taipa, Macau SAR, China
Institutions: Pennsylvania State University (United States); University of Macau (Macao SAR China)
Journal: Neurobiology of language (Cambridge, Mass.), volume 7, article NOL.a.245
Dates: received 31 July 2025; accepted 11 February 2026; published online 23 April 2026
Type: Research article · Language: English
License: CC BY
Identifiers: DOI 10.1162/nol.a.245 · PMID 42088907 · PMCID PMC13137885 · OpenAlex W7131129959
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: fMRI (modality), cognitive (subfield)
Methods: Spectral & time-frequency, Connectivity, Statistics, fMRI & imaging, Physiology & signal measures
Keywords: aging, disfluency, functional magnetic resonance imaging (fMRI), resting-state functional connectivity (RSFC), speech production
Topic: Neurobiology of Language and Bilingualism (Cognitive Neuroscience, Neuroscience), according to OpenAlex
Funding: NIA NIH HHS (R01 AG034138, T32 AG049676)
Citations: not cited yet (Europe PMC); 99 references in the paper

Abstract

Fluent speech production remains largely preserved across adulthood, yet subtle disruptions such as pauses, repetitions, and revisions become more common with age. These disfluencies may reflect underlying cognitive and neural changes that accompany aging, particularly in executive function (EF) and large-scale brain network organization. In this study, we examined whether EF and resting-state functional connectivity (RSFC) independently or jointly explained age-related differences in naturalistic speech disfluencies in an adult lifespan sample (n = 252, ages 20–81 years). RSFC was used to assess network segregation within three systems implicated in language and cognitive control: language network, default mode network (DMN), and multiple demand (MD) network. These task-independent connectivity patterns provide insight into how the brain’s functional architecture impacts speech production and its age-related vulnerabilities. Our findings indicate that age was associated with increased rates of specific disfluency subtypes, such as unfilled pauses, repetitions, and revisions, as well as lower EF and lower language, MD, and DMN network segregation. Although increasing age was associated with lower EF, EF performance did not predict disfluencies or mediate their age-related increase. In contrast, higher DMN segregation predicted lower overall disfluencies, repetitions, and revisions. Age moderated the relationship between DMN segregation and repetitions, with a significant association only in younger and middle-aged adults, suggesting weaker brain–behavior relationships at older ages. DMN segregation also partially mediated the relationship between age and revisions. These findings suggest that while EF relates to planning-related disruptions, changes in functional brain organization may more directly contribute to age-related increases in self-monitoring disfluencies.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 3 matches between paragraphs and lines of code.

OSF vp9za

License: none: the authors keep all their rights
State: the link answers, verified on 29 September 2026
Evidence: files inventoried
Languages: R (1)
Size: 8 files, 1 script
Software Heritage: not checked
Found in: “Data Availability Statement”
Holds: 1 notebook
Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation
Tools: broom (1 file), ggplot2 (1 file), psych (1 file), tidyverse (1 file)
Availability: 1 check, the latest on 29 September 2026: the link answers (HTTP 200)
  • 29 September 2026: the link answers (HTTP 200)
1 file
At the source: osf.io/vp9za

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 1 script, each with its path and the digest of its content;
  • 3 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data Availability Statement

The data and analysis scripts are openly available on the OSF at: https://osf.io/vp9za. Raw imaging data may be available upon request.

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 29 September 2026: the first record

Recorded: type, language, journal, volume, pages, dates, 3 authors, 5 keywords, 1 funder, 91 references.

Cite

This paper

Nakamura, M. S., Zhang, H., & Diaz, M. T. (2026). Age-Related Differences in Resting-State Functional Connectivity Predict Specific Patterns of Speech Disfluency. Neurobiology of language (Cambridge, Mass.), 7, NOL.a.245. https://doi.org/10.1162/nol.a.245

BibTeX

@article{nakamura2026age,
author = {Nakamura, Megan S and Zhang, Haoyun and Diaz, Michele T},
title = {{Age-Related Differences in Resting-State Functional Connectivity Predict Specific Patterns of Speech Disfluency}},
journal = {Neurobiology of language (Cambridge, Mass.)},
year = {2026},
month = apr,
volume = {7},
pages = {NOL.a.245},
publisher = {MIT Press},
issn = {2641-4368},
doi = {10.1162/nol.a.245},
url = {https://doi.org/10.1162/nol.a.245},
pmid = {42088907},
pmcid = {PMC13137885}
}

RIS

TY - JOUR
AU - Nakamura, Megan S
AU - Zhang, Haoyun
AU - Diaz, Michele T
TI - Age-Related Differences in Resting-State Functional Connectivity Predict Specific Patterns of Speech Disfluency
T2 - Neurobiology of language (Cambridge, Mass.)
J2 - Neurobiol Lang (Camb)
PY - 2026
DA - 2026/04/23
VL - 7
SP - NOL.a.245
SN - 2641-4368
PB - MIT Press
DO - 10.1162/nol.a.245
UR - https://doi.org/10.1162/nol.a.245
LA - en
ER -

CSL-JSON

{
"id": "10.1162/nol.a.245",
"type": "article-journal",
"title": "Age-Related Differences in Resting-State Functional Connectivity Predict Specific Patterns of Speech Disfluency",
"container-title": "Neurobiology of language (Cambridge, Mass.)",
"author": [
{
"family": "Nakamura",
"given": "Megan S"
},
{
"family": "Zhang",
"given": "Haoyun"
},
{
"family": "Diaz",
"given": "Michele T"
}
],
"container-title-short": "Neurobiol Lang (Camb)",
"volume": "7",
"page": "NOL.a.245",
"DOI": "10.1162/nol.a.245",
"PMID": "42088907",
"PMCID": "PMC13137885",
"ISSN": "2641-4368",
"publisher": "MIT Press",
"URL": "https://doi.org/10.1162/nol.a.245",
"language": "en",
"issued": {
"date-parts": [
[
2026,
4,
23
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1038/s41467-026-76598-x
Preserved topography, lateralization, selectivity, and functional connectivity of the language network in older brains.
Journal: Nature communications
In common: fMRI, cognitive, 12 references
[2] doi:10.1038/s42003-026-10282-0 [code]
Genetic risk of Alzheimer's disease is associated with loss of brain network segregation in midlife.
Journal: Communications biology
In common: psych, broom, ggplot2, 1 other tool, 5 references
[3] doi:10.1016/j.neuroimage.2026.122186
Aging, brain network connectivity, and cognition: evidence from resting-state dynamic arterial spin labeling.
Journal: NeuroImage
In common: fMRI, cognitive, 7 references
[4] doi:10.1162/imag.a.1285 [code]
Individualized mapping of functional brain networks in older adulthood.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: fMRI, 7 references
[5] doi:10.1186/s40337-026-01671-1 [code]
Aberrant large- and mesoscale network segregation and integration in bulimia nervosa.
Journal: Journal of eating disorders
In common: broom, ggplot2, tidyverse, fMRI, 4 references
[6] doi:10.1162/netn.a.542
High brain network system segregation is differentially linked with cognitive performance across the life span.
Journal: Network neuroscience (Cambridge, Mass.)
In common: fMRI, cognitive, 6 references
[7] doi:10.1162/imag.a.1283 [code]
The language network responds robustly to sentences across tasks.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: broom, tidyverse, fMRI, 4 references
[8] doi:10.1162/imag.a.1338 [code]
Systematic fMRI signal differences across cohorts alter lifespan trajectories of functional brain networks.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: fMRI, 5 references
[9] doi:10.1093/braincomms/fcag279 [code]
Network flexibility facilitates treatment-induced recovery in post-stroke aphasia.
Journal: Brain communications
In common: psych, broom, ggplot2, 1 other tool, fMRI, 1 reference
[10] doi:10.1162/imag.a.1298 [code]
Frontoparietal control-default mode connectivity predicts TMS effects on cognitive control.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: psych, ggplot2, tidyverse, cognitive, 2 references

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.