RMET behavioral analyses (fMRI version)

Summary

FYI:

• Included here are n = 69 with final group variable (actual n = 73 w/scan data to date).
• n = 3 with accuracy <50% (social trials only)
• Zoe identified that all 3 speak English as a second language, though their English ability was sufficient for their enrollment in the study and they all learned English during childhood. One of them also has the highest WTAR score (120) of the n = 19 ESL participants, so unclear whether English proficiency is the sole reason for their low accuracy…

Results

Note: conclusions not based on adjusted p value for multiple comparisons, I just wanted to get a sense of where the data were for now.

• Effect of trial type on performance (see boxplots):
  • Mean RTs on non-social trials are faster than on social trials, across all 3 groups.
  • Accuracy on non-social trials is higher than on social trials, across all 3 groups.
• Effect of group on performance (see boxplots):
  • Non-social trials: no major group differences in mean RT or accuracy.
  • Social trials:
    • The Low-exposed group is slower than either the PTSD or Resilient group.
    • The PTSD group is less accurate than either the Low-exposed or Resilient group.
• Effect of MOS and/or group on performance (see scatter- and boxplots):
  • Scatterplots show the linear association between RMET performance and MOS scores, within each group and trial type (no statistical tests here, I did run the linear models for the three-way interaction of group x trialtype x MOS subscale score but I was concerned about overfitting because there are so many contrasts we are looking at relative to the sample size).
• Unsurprisingly…across all MOS subscales, PTSD group scores are significantly lower than either Resilient or Low-exposed groups. Resilient and Low-Exposed groups reported similar levels of social support (see boxplots).

[1] 364.2093
[1] 2982.781
`summarise()` has grouped output by 'bids_id', 'run'. You can override using the `.groups` argument.

Descriptives by group

`summarise()` has grouped output by 'trialtype', 'group'. You can override using the `.groups` argument.

`summarise()` has grouped output by 'trialtype'. You can override using the `.groups` argument.

Subjects with <50% accuracy

sub-025 = NWTC-033 (Spanish 1st language, learned English age 6, grad. 4 year college)

sub-035 = NWTC-032 (Spanish 1st language, learned English age 5, grad. 4 year college)

sub-039 = NWTC-036 (German 1st language, learned English age 4, grad. 4 year college)

Boxplots: Run & trial type

Run 1 vs. run 2

Call:
lm(formula = meanRT ~ trialtype, data = rmetDat)

Residuals:
    Min      1Q  Median      3Q     Max 
-538.45 -180.91   -0.76  165.42  871.20 

Coefficients:
                Estimate Std. Error t value             Pr(>|t|)    
(Intercept)      1258.68      21.70  58.009 < 0.0000000000000002 ***
trialtypesocial   226.17      30.69   7.371     0.00000000000215 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 251.2 on 266 degrees of freedom
Multiple R-squared:  0.1696,    Adjusted R-squared:  0.1665 
F-statistic: 54.33 on 1 and 266 DF,  p-value: 0.00000000000215


Call:
lm(formula = accuracy ~ trialtype, data = rmetDat)

Residuals:
     Min       1Q   Median       3Q      Max 
-0.29068 -0.06656  0.01978  0.06227  0.24200 

Coefficients:
                 Estimate Std. Error t value            Pr(>|t|)    
(Intercept)      0.899899   0.008235  109.28 <0.0000000000000002 ***
trialtypesocial -0.197458   0.011646  -16.95 <0.0000000000000002 ***
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.09533 on 266 degrees of freedom
Multiple R-squared:  0.5194,    Adjusted R-squared:  0.5176 
F-statistic: 287.5 on 1 and 266 DF,  p-value: < 0.00000000000000022

Boxplots: Group comparison of RMET performance

MOS Scoring

data<-read.csv("/Users/sarenseeley/Dropbox/Postdoc/nwtc_study/meetings/nwtc_data_cleaned_forMeeting_04-06-22.csv", strip.white=FALSE, na.strings="NA")

# get NWTC to BIDS ID conversion
bids_key <- read.csv("/Users/sarenseeley/Dropbox/Postdoc/nwtc_study/notes/nwtc_bids_key.csv")[,-3]

# for people with a 3-month gap between V1 and V2:
mos_int <- data %>% filter(record_id=="NWTC-023" | record_id=="NWTC-025" | record_id=="NWTC-032" | record_id=="NWTC-036" | record_id=="NWTC-079") %>% select(starts_with("mos_") & ends_with("v4_3m"))

ids <- data %>% filter(record_id=="NWTC-023" | record_id=="NWTC-025" | record_id=="NWTC-032" | record_id=="NWTC-036" | record_id=="NWTC-079") %>% select(record_id)
colnames(mos_int)<-gsub("_v4_3m", ".use", names(mos_int)) # rename vars to indicate these are the ones to use
mos_int <- bind_cols(ids,mos_int)

# for people without an interval gap:
mos <- data %>% filter(!record_id %in% mos_int$record_id) %>% select(starts_with("mos_") & !ends_with("_v4_3m"))
colnames(mos) <- paste(colnames(mos),"use",sep=".")
ids <- data %>% filter(!record_id %in% mos_int$record_id) %>% select("record_id")
mos <- bind_cols(ids,mos)



# bind rows, score, and merge back into data
mos <- bind_rows(mos, mos_int)
mos <- left_join(mos,bids_key, by = "record_id") %>% 
                 mutate(tot_mos_emoSup = rowMeans(select(., c(mos_3.use,
                                                      mos_4.use,
                                                      mos_8.use,
                                                      mos_9.use,
                                                      mos_13.use,
                                                      mos_16.use,
                                                      mos_17.use,
                                                      mos_19.use)), na.rm=TRUE),
                        tot_mos_tanSup = rowMeans(select(., c(mos_2.use,
                                                      mos_5.use,
                                                      mos_12.use,
                                                      mos_15.use)),na.rm=TRUE),
                        tot_mos_affSup = rowMeans(select(., c(mos_6.use,
                                                      mos_10.use,
                                                      mos_20.use)),na.rm=TRUE),
                        tot_mos_posSup = rowMeans(select(., c(mos_7.use,
                                                      mos_11.use,
                                                      mos_18.use)),na.rm=TRUE),
                        tot_mos_overallSup = rowMeans(select(., c(mos_2.use,
                                                          mos_3.use,
                                                          mos_4.use,
                                                          mos_5.use,
                                                          mos_6.use,
                                                          mos_7.use,
                                                          mos_8.use,
                                                          mos_9.use,
                                                          mos_10.use,
                                                          mos_11.use,
                                                          mos_12.use,
                                                          mos_13.use,
                                                          mos_14.use,
                                                          mos_15.use,
                                                          mos_16.use,
                                                          mos_17.use,
                                                          mos_18.use,
                                                          mos_19.use,
                                                          mos_20.use)),na.rm=TRUE),
                        tot_mos_nClose = mos_1.use)

mos <- mos %>% mutate(tot_mos_emoSup = 100*((tot_mos_emoSup-1)/4),   # convert to 1-100 scale
                      tot_mos_tanSup = 100*((tot_mos_tanSup-1)/4),
                      tot_mos_affSup = 100*((tot_mos_affSup-1)/4),
                      tot_mos_posSup = 100*((tot_mos_posSup-1)/4),
                      tot_mos_overallSup = 100*((tot_mos_overallSup-1)/4),
)
  

data <- left_join(mos,data,by="record_id")
tot_mos <- mos %>% select(starts_with("tot") | starts_with("bids"))
rmetWide <- left_join(rmetWide,tot_mos, by ="bids_id")
rmetLong <- left_join(rmetLong,tot_mos, by ="bids_id")
rmetDat <- left_join(rmetDat,tot_mos, by ="bids_id")

# checking MOS scores

mos_dayle <- read.csv("/Users/sarenseeley/Dropbox/Postdoc/nwtc_study/meetings/additional_scales_for_SNT_analyses.csv", strip.white=FALSE, na.strings="NA")[,1:25]

mos_dayle <- mos_dayle %>% mutate(source = "dayle") %>% rename(tot_mos_overallSup = overall_support,
                                                              tot_mos_emoSup =  emotional_informational,
                                                               tot_mos_tanSup = tangible,
                                                               tot_mos_affSup = affectionate,
                                                              tot_mos_posSup = pos_social_interaction)
mos <- mos %>% mutate(source="saren")


mos1 <- mos %>% select(c("record_id", starts_with("tot"), "source") & !"tot_mos_nClose") %>% select(sort(names(.))) %>% 
  mutate(tot_mos_overallSup = round(tot_mos_overallSup,0),
         tot_mos_emoSup = round(tot_mos_emoSup,0),
         tot_mos_tanSup = round(tot_mos_tanSup,0),
         tot_mos_affSup = round(tot_mos_affSup,0),
         tot_mos_posSup = round(tot_mos_posSup,0))
                        

mos_dayle1 <- mos_dayle %>% select(c("record_id", starts_with("tot"), "source")) %>% select(sort(names(.)))

mos_check <- bind_rows(mos1,mos_dayle1) %>% filter(!record_id %in% c("NWTC-023","NWTC-025","NWTC-032","NWTC-036","NWTC-079"))
mos_check <- pivot_wider(mos_check, id_cols = "record_id",
                         names_from = starts_with("source"),
                         values_from = starts_with("tot_mos")) 

mos_check <- mos_check %>% filter(!is.na(tot_mos_emoSup_saren) & !is.na(tot_mos_emoSup_dayle)) %>%
 mutate(emo_match = as.factor(ifelse(tot_mos_emoSup_saren == tot_mos_emoSup_dayle, 1, 0)),
        aff_match = as.factor(ifelse(tot_mos_affSup_saren == tot_mos_affSup_dayle, 1, 0)),
        tan_match = as.factor(ifelse(tot_mos_tanSup_saren == tot_mos_tanSup_dayle, 1, 0)),
        pos_match = as.factor(ifelse(tot_mos_posSup_saren == tot_mos_posSup_dayle, 1, 0)),
        overall_match = as.factor(ifelse(tot_mos_overallSup_saren == tot_mos_overallSup_dayle, 1, 0)))

        
write.csv(mos_check,"/Users/sarenseeley/Dropbox/Postdoc/nwtc_study/meetings/mos_scoring_comparison.csv")

Scatterplots: Mean RT relationship with MOS scores

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'
Warning: Removed 4 rows containing non-finite values (stat_smooth).
Warning: Removed 4 rows containing missing values (geom_point).

Scatterplots: Accuracy relationship with MOS scores

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'

`geom_smooth()` using formula 'y ~ x'
Warning: Removed 4 rows containing non-finite values (stat_smooth).
Warning: Removed 4 rows containing missing values (geom_point).

Boxplots: Group differences in MOS social support

Warning: Removed 1 rows containing non-finite values (stat_boxplot).
Warning: Removed 1 rows containing non-finite values (stat_signif).

Frequencies of words incorrect/missed