Depressive Sx and alterations in event duration at grade reveal

Baseline Model (lmer)

formula(dur.dep.lmer)
## current_duration ~ z_PHQ * current_PE + (1 | subject)

Effect of PHQ on Duration:

unclear in 2-way interaction

# CELL: plot simple moderation effect
plot_model(dur.dep.lmer, type = "pred", terms = c("current_PE[-20,20]","z_PHQ[0,1.5]"))

Follow-up: raw data plot

p_center <- plot_duration_by_PE_bins(
  recur_counts_psy,
  pe_min              = -20,
  pe_max              =  20,
  bin_width           = 10,
  center_bin_halfwidth= 2,
  min_n_per_bin       = 15,
  TR_sec              = 2,
  highlight_sep       = TRUE
) +
  geom_vline(xintercept = 0, linetype = "dashed", alpha = 0.4)

plot(p_center)

Duration alterations in higher-PHQ pts seem to differ by PE valence

suggests we should isolate PE sign and magnitude in stats models…

Refactored model (brms)

formula(fit_dur_PHQ_absPE_sign)
## duration_s ~ z_PHQ * z_abs_PE * PE_sign + (1 | subject)
family = gaussian()

Credible differences are limited to positive PE events

plot(bp3)

Updated Conclusion:

Higher depressive symptoms linked to lower duration events after positive PEs