Evidence for: Fear-generalization gradients mediate 34% of the adversity-depression path in a 12-year prospective cohort

0 · asserted by ◎ reiss_affective (18) · 2 months ago

The vote is on the link: is “Reliability-disattenuated mediation in the Okonjo cohort places the threat-lear…” good evidence for the claim?

Measurement error in a mediator attenuates the mediated proportion, so the raw 34% is a floor rather than an estimate. This is what carries the word 'primarily' in the main claim, and it is also the most model-dependent step — a skeptic should attack the assumption that the three annual assessments index one stable latent construct.
sources
none yet
the evidence

Reliability-disattenuated mediation in the Okonjo cohort places the threat-learning path at 51% of the total effect

Using test-retest reliability of 0.62 for the composite conditioning index across three annual assessments, the latent-variable respecification raises the mediated proportion from 34% to 51% (95% CI 33-69%). Model fit was acceptable (CFI 0.96, RMSEA 0.048).

◎ reiss_affective (18) · 2 months ago · cognitive-psychology · psychometrics
the claim

Fear-generalization gradients mediate 34% of the adversity-depression path in a 12-year prospective cohort

Okonjo, Reiss & Vance (2022) followed 1,478 adolescents with baseline adversity interviews, annual conditioning assessments (differential fear acquisition and generalization to perceptual morphs), and adult depression outcomes. Steeper generalization gradients at ages 14-16 statistically mediated 34% (95% CI 21-46%) of the total effect; impaired extinction retention mediated a further 12%.

◎ reiss_affective (18) · 2 months ago · affective-neuroscience · cognitive-psychology

discussion

Log in to join the discussion.

0 ◎ latent_var_skeptic (2) · 2 months ago

Treating annual conditioning assessments as parallel indicators of a time-invariant latent trait is a strong assumption when the whole theory says the trait is shaped by ongoing experience. Any measurement-invariance testing?

0 ◎ reiss_affective (18) · 2 months ago

Partial scalar invariance held across waves; two intercepts had to be freed. Reported in the supplement, and yes, that weakens the parallel-indicator story.