As Predicted:Individual differences in fluid ability (#7195)
Created: 12/06/2017 06:41 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 availble 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?
Can a watershed model explain individual differences in fluid ability in children and adolescents?
- The watershed model (Kievit et al., 2016) predicts a hierarchical relationship between white matter microstructure, endophenotypes (such as working memory and processing speed) and fluid ability (gf).3) Describe the key dependent variable(s) specifying how they will be measured.
- Neurally, we will use tract-based mean FA in 10 JHU white matter tracts.
- Behaviourally, we will use raw scores on tests of gf (WASI matrix reasoning), working memory (Automated Working Memory Battery) and processing speed (TEACH RBBS RT, DKEFS trails motor speed, PHAB rapid naming).
- We will also include age as a covariate to examine model heterogeneity (i.e. age differentiation)4) How many and which conditions will participants be assigned to?
N/A5) Specify exactly which analyses you will conduct to examine the main question/hypothesis.
We will use multivariate Structural Equation Models (SEMs) to answer our research question. These models will be implemented using Full Information Maximum Likelihood with Satorra-Bentler scaled test statistics in R's lavaan for hypothesis 1-4 and OpenMX for hypothesis 5.
Hypothesis 1. Working memory, gf and processing speed are separable constructs.
- We will compare a 3-factor model (working memory, gf and speed as separate factors) to 2-factor models (A: working memory/gf as a single factor, separate from speed; B: speed/gf as a single factor, separate from working memory), to a single-factor model (working memory/gf/speed as a single factor). We expect the 3-factor model to fit best.
Hypothesis 2. Individual differences in gf are predicted by working memory and processing speed.
- We will use a Multiple Indicator, Multiple Cause (MIMIC) models, which will allow us to compare a full model in which paths between working memory and speed and gf are freely estimated to models in which either of these paths are constrained to 0. We expect the full MIMIC model to fit best.
Hypothesis 3. White matter microstructure is a multi-dimensional construct
- We will examine global measures of a single-factor model of white matter, and expect that a single factor model will show poor fit, as in Kievit et al. 2016.
Hypothesis 4. There is a hierarchical relationship between white matter microstructure, endophenotypes (working memory, processing speed) and gf (e.g. Fry and Hale, 1996; Tourva et al., 2016; Kail et al., 2015; Anderson, 2017; Kievit et al., 2016)
- We will examine the fit of this watershed model and compare it to constrained and alternative models (see Kievit et al., 2016, Figure 6 for details of the model). We hypothesize the watershed model will outperform alternative candidates, as judged by LRT and AIC tests
Hypothesis 5. The contribution of working memory and processing speed to gf changes with age (e.g. Demetriou, 2014).
- We will identify changes in the relationship between gf, endophenotypes and white matter with age in a data-driven way (Brandmaier et al., 2016). We will use SEM trees with age as a covariate, and expect the parameters to either remain static with age or decrease.6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations.
Exclusion criteria for referrals to the CALM clinic, from which our sample stems, were: signiﬁcant and uncorrected known problems in vision or hearing and a native language other than English. Other than that, we will not impose any exclusion criteria on individual data points.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.
We will use the CALM sample it its N = 550 release. See http://calm.mrc-cbu.cam.ac.uk/. Additionally, we are aiming to validate our models in two other cohorts - the NKI Rockland Sample (http://fcon_1000.projects.nitrc.org/indi/enhanced/) and the ABCD sample (https://addictionresearch.nih.gov/abcd-study). For both of these samples we will use the data available in January 2018.8) Anything else you would like to pre-register?
(e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?)
All of the data we are planning to use stems from existing, publicly available datasets with timelined data release approvals along the way. The study was preregistered here before full data access to any of the samples had been received (i.e. no data inspection or analysis has taken place prior to this preregistration).