Spss statistics v23

The study datasets and code used to generate the results is available for IBM SPSS Statistics v23 and StataCorp LP Stata/IC 13.1 upon request from the author.

Statistical analyses were carried out with IBM SPSS Statistics v23 and Stata/IC 16.1. Univariate analysis of variance was used for calculating the power of the sample and probability of type II error. Pearson correlation was used to test bivariate association with all variables. Point biserial correlation was also carried out to account for binary variables (coded as exposed – 1 or non-exposed – 0) that yielded the same results as Pearson. Second, linear regression analysis and stepwise backward regression were conducted with variables of the four predictor domains (demographic, perinatal, neonatal and developmental characteristics) for the outcome variable. Subscales of the BSID-III were tested separately, and also as aggregate measure. Significance level was set at alpha = 0.05. The models were evaluated by explained variance (R²), measure of residuals (root mean square error), and measure of the models’ fit (Akaike’s information criterion). Third, predictors of the best fitting model were identified by stepwise backward regression analysis, setting alpha for the removal of predictors at 0.05.