. 2019 Dec 18;9:19399. doi: 10.1038/s41598-019-55350-0

Table 1.

A linear analysis replicates analysis of variance results.

	Estimate	Standard Error^a	95% CI	p
a. Mnemonic Discrimination: Same-race vs Other-race faces	0.22	0.05	(0.13, 0.31)	<0.0001
b. Match-to-Sample: Same-race vs Other-race faces	−0.02	0.07	(−0.16, 0.12)	0.7308
c. Interaction between task and race	0.25	0.09	(0.08, 0.42)	<0.005

Estimate

Standard Error^a

95% CI

a. Mnemonic Discrimination:

Same-race vs Other-race faces

0.22

0.05

(0.13, 0.31)

<0.0001

b. Match-to-Sample:

Same-race vs Other-race faces

−0.02

0.07

(−0.16, 0.12)

0.7308

c. Interaction between task and race

0.25

0.09

(0.08, 0.42)

<0.005

A linear analysis was run to account for differences that may be attributable to unmatched sample sizes. The model was fit using generalized estimating equations, which are robust to sample size differences across groups. d’ was modeled as a linear combination of race, task, and the interaction of the two. A and B: The analysis reveals significant differences in estimated population means for SR and OR faces in the mnemonic discrimination but not match-to-sample task. C. There is a significant interaction between task and race with the difference in the estimated d’ between participant’s recognition of SR and OR faces 0.25 larger in the mnemonic discrimination task than the difference in the estimated d’ between SR and OR faces in the match-to-sample task.

^aHeteroscedasticity-consistent “sandwich” standard errors are used to allow for differences in the variance of model errors across different participant subgroups.