Set up multiple imputation

[conig]

We will analyse all participants who provide outcome data for a health outcome. Where there is missing predictor or covariate data for an individual, we will assume that it to be missing at random or completely at random. We will impute these missing values using multiple imputation by chained equations using 50 imputed datasets. We will use Rubin’s rule to combine the results from all imputed datasets into one single set of results.

[conig]

We are abandoning multiple imputation for now as we're moving over to lavaan and latent growth curve models.