Enhancing Within-Person Estimation of Neurocognition and the Prediction of Externalizing Behaviors in Adolescents

Sam Paskewitz; Inti A. Brazil; Ilker Yildirim; Sonia Ruiz; Arielle Baskin-Sommers

doi:10.5334/cpsy.112

Figures & Tables

Illustration of LPA with too few or too many profiles.

Table 1

Variables.

(A) NEUROCOGNITIVE (INDICATOR) VARIABLES (x)
COGNITIVE PROCESS	TASK	DESCRIPTION
Executive Function
cognitive/attentional control	Flanker (NIH Toolbox)	uncorrected standard score
processing speed	Pattern Comparison (NIH Toolbox)	uncorrected standard score
response inhibition	Stop Signal	–1× stop signal reaction time
working memory	Emotional N-Back	proportion correct on two-back trials
Learning and Memory
visual-spatial memory	Picture Sequence Memory (NIH Toolbox)	uncorrected standard score
auditory memory	Rey Auditory Verbal Learning	immediate recall total correct
recognition memory	Emotional N-Back	recognition trials d′
General Cognition
vocabulary	Picture Vocabulary (NIH Toolbox)	uncorrected standard score
reading decoding	Oral Reading Recognition (NIH Toolbox)	uncorrected standard score
spatial processing	Little Man Task	% correct/avg. correct response time
decision-making	Game of Dice	# safe choices – # risky choices
(B) OUTCOME VARIABLES (y).
VARIABLE	SAMPLE SIZE (2-YEAR FOLLOW-UP)	SAMPLE SIZE (3-YEAR FOLLOW-UP)
externalizing (CBCL)	4977	4742
rule-breaking (CBCL)	4977	4742
aggression (CBCL)	4977	4742
internalizing (CBCL)	4977	4742
positive urgency (UPPS-P)	5230	–
negative urgency (UPPS-P)	5230	–

DPM-LPA prior distribution (log-scale) on participant latent profile membership (z_1:n) as a function of the number of latent profiles. Calculated with n = 5000 participants, with each new profile created by splitting the last one in half.

Comparisons of conventional LPA models with differing numbers of profiles (all variables are coded so that larger means better). Where applicable, a dashed horizontal line indicates the corresponding value from DPM-LPA.

Estimated latent profile means (μ) from DPM-LPA.

Estimated DPM-LPA latent profile outcome means (95% posterior credible interval).

Table 2

Summary of outcome analysis results. log₁₀ (BF10) is the Bayes factor for the Bayesian ANOVA on a base-10 logarithmic scale. r² is the effect size. $μ_{j}^{(y)}$ is the mean value of the relevant outcome variable for profile j.

(A) TWO YEAR FOLLOW-UP.
VARIABLE	log10 (BF₁₀)	r²	POST-HOC RESULTS
externalizing	3.51	0.0090	$μ_{2}^{(y)} = μ_{4}^{(y)} < μ_{1}^{(y)} = μ_{6}^{(y)} = μ_{7}^{(y)} < μ_{3}^{(y)} = μ_{5}^{(y)} = μ_{8}^{(y)} = μ_{9}^{(y)}$
rule-breaking	11.59	0.0167	$μ_{2}^{(y)} = μ_{4}^{(y)} < μ_{1}^{(y)} = μ_{7}^{(y)} < μ_{3}^{(y)} = μ_{5}^{(y)} = μ_{6}^{(y)} = μ_{8}^{(y)} = μ_{9}^{(y)}$
aggression	1.22	0.0070	$μ_{2}^{(y)} = μ_{4}^{(y)} < μ_{1}^{(y)} = μ_{6}^{(y)} = μ_{7}^{(y)} < μ_{3}^{(y)} = μ_{5}^{(y)} = μ_{8}^{(y)} < μ_{9}^{(y)}$
positive urgency	18.87	0.0230	$μ_{2}^{(y)} = μ_{4}^{(y)} < μ_{1}^{(y)} = μ_{7}^{(y)} < μ_{3}^{(y)} = μ_{5}^{(y)} < μ_{6}^{(y)} = μ_{8}^{(y)} = μ_{9}^{(y)}$
negative urgency	–0.47	0.0052	NA (inconclusive)
internalizing	–1.57	NA	NA (no differences)
(B) THREE YEAR FOLLOW-UP.
VARIABLE	log10 (BF₁₀)	r²	POST-HOC RESULTS
externalizing	2.11	0.0079	$μ_{2}^{(y)} = μ_{4}^{(y)} = μ_{7}^{(y)} < μ_{1}^{(y)} = μ_{3}^{(y)} < μ_{5}^{(y)} = μ_{6}^{(y)} = μ_{8}^{(y)} = μ_{9}^{(y)}$
rule-breaking	8.01	0.0140	$μ_{2}^{(y)} = μ_{4}^{(y)} < μ_{1}^{(y)} = μ_{7}^{(y)} < μ_{3}^{(y)} = μ_{5}^{(y)} = μ_{6}^{(y)} = μ_{8}^{(y)} = μ_{9}^{(y)}$
aggression	–0.85	NA	NA (no differences)
internalizing	–0.62	NA	NA (no differences)

Enhancing Within-Person Estimation of Neurocognition and the Prediction of Externalizing Behaviors in Adolescents

Figures & Tables

Figure 1

Figure 2

Table 1

Figure 3

Figure 4

Figure 5

Figure 6

Table 2

Paradigm

My account