These main effects remained significant in an ANCOVA model including the following as covariates: prior programming and robotics competition experience, F 1, 474) = 89.61, p < .001, η p 2 = .16; robotics and programming interest, F 1, 476) = 11.52, p < .001, η p 2 = .02; and both experience and interest covariates, F 1, 466) = 13.82, p < .001, η p 2 = .03.03
For Level 2 hospitals, the main effects remained essentially unchanged following exclusion of those patients with hospital re-admissions; however the interaction between health insurance and income became non-significant.
After stratifying by response to both of the blinding questions, the overall conclusion on the main effects remained the same (statistical reviewer effect = 5.9, 95% CI = 3.4 to 8.5; checklist effect = −1.43, 95% CI = −3.8 to 0.99).
If an interaction term was significant the corresponding main effects remained in the model, whether or not they were significant.
Restricted analyses were also undertaken after removing the 12 infants born before 37 weeks' gestation; main effects remained unchanged from those reported here for the whole cohort.
The inclusion of the interaction variables with risk status did not affect the overall model as depicted in Figure 2, as all the same main effects remained significant at a similar level.
The main effects remain stable when including interaction terms into the model.
Throughout the analysis, we report robust standard errors, but the main effects remain significant also when standard errors are clustered at the municipal level.
The pattern of change in the sub-cohort main effect remained unaffected by the addition of the nested air lead variable.
Finally, there were clear differences between the genotypes in the levels of Aβ42, which were highly significant (P = 0.0034), although when Aβ42 was fitted as a covariate, the main effect remained (P = 0.0157).
However, the main air pollution effects remained stable in the analyses for seven of the eight covariates except age in our models.
