'Predicting cognitive resilience among low-income children from rsfMRI'
Created: 06/13/2019 06:54 AM (PT)
This is an anonymized version of the pre-registration. It was created by the author(s) to use during peer-review.
A non-anonymized version (containing author names) should be made available by the authors when the work it supports is made public.
1) Have any data been collected for this study already?
It's complicated. We have already collected some data but explain in Question 8 why readers may consider this a valid pre-registration nevertheless.2) What's the main question being asked or hypothesis being tested in this study?
What are the neural indicators of cognitive “resilience” among children living below the poverty line? We hypothesize that stronger lateral frontoparietal network (LFPN) within-network connectivity and weaker LFPN-default mode network (DMN) between-network connectivity will be associated with higher cognitive test scores, based on previous research with higher-SES children. In addition, we hypothesize that stronger left rostrolateral prefrontal cortex (RLPFC)-inferior parietal lobule (IPL) connectivity will be associated with higher reasoning scores. If we do not find these expected links, it may be because lower-income children’s cognitive test scores are more strongly influenced by socioemotional factors, and/or because the analysis pipeline obscures important individual differences in network topography.3) Describe the key dependent variable(s) specifying how they will be measured.
Cognitive scores: Z-score and sum of Flanker, DCCS, and WISC matrix reasoning at T1 and T2
Reasoning alone: WISC matrix reasoning at T1 and T2
Resting state connectivity (average strength of correlation across nodes within/between specific brain networks/nodes):
1) within LFPN, as identified using the ABCD in-house processing
2) between LFPN and DMN, as identified using the ABCD in-house processing
3) between left RLPFC and left IPL, using ROIs from the Destrieux atlas, identifying regions closest to those used in Wendelken et al., 20154) How many and which conditions will participants be assigned to?
None.5) Specify exactly which analyses you will conduct to examine the main question/hypothesis.
All linear models below will be performed using cross-validation, iteratively leaving one-fifth of the data out and re-assessing model performance. Significance will be tested by comparing models with the IV of interest to those without using the anova function in R.
1a) Linear model predicting cognitive scores at T1 and T2 (separately) from LFPN within-network connectivity and LFPN-DMN between-network connectivity (separately).
1b) If LFPN connectivity at T1 predicts cognitive scores at T2, test whether this holds when covarying for cognitive scores at T1.
2a) Linear model predicting reasoning at T1 and T2 (separately) from left RLPFC-IPL connectivity.
2b) If left RLPFC-IPL at T1 predicts reasoning at T2, test whether this holds when covarying for cognitive scores at T1.
3) If analyses above yield significant results, test whether resting state metrics predict cognitive scores/reasoning above and beyond potential buffering factors, such as parental years of education.6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations.
We will exclude children who do not fall below 0 on the income-to-needs ratio, as calculated using the supplemental poverty measure for their specific region and the number of people in their home. For children who do not provide data on all three cognitive tests, we will include them only in secondary analyses of reasoning. In addition, we will use ABCD’s usability guidelines to determine usability of fMRI data.7) How many observations will be collected or what will determine sample size?
No need to justify decision, but be precise about exactly how the number will be determined.
As many as usable at T1 (baseline) and T2 (one-year follow-up).8) Anything else you would like to pre-register?
(e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?)
Data have already been collected but we have not yet obtained access to them.
We may redo resting state processing with an in-house pipeline; this would allow us to
- omit global signal regression, which is not advised for detection of anti-correlations
- get more precise ROI estimates for our particular needs, and
- account for inter-individual variability in network topography. Specifically:
1) LFPN would be identified using a DLPFC seed from Fox et al., 2005*
2) DMN would be identified using an MPFC seed from Fox et al., 2005*
3) RLPFC and IPL would be identified using seeds from Wendelken et al., 2015
* The resulting networks would be visually inspected to ensure that they capture canonical LFPN and DMN regions in this population; we will take an alternate approach (e.g., Kong et al., Cerebral Cortex, 2019), to identify these.