Phenotype and covariate files should be in PLINK format. The first two columns should provide sample IDs, with subsequent columns providing values for each phenotype or covariate. The files can contain a header row, but then the first two elements must be named either FID & IID or ID1 & ID2. When a phenotype file contains more than one phenotype (i.e., more than three columns), you should specify which to analyse using the option --mpheno <integer> (to specify by number) or --pheno-name <name> (to specify by name). Alternatively, some functions (most notably REML) allow for testing of all phenotypes by adding --mpheno -1. There is no option to select covariates (all covariates will be used, so to use only a subset, you should first make a reduced covariate file).
Missing values should be denoted by NA. Note that whereas PLINK also treats -9 as missing, this is not the case in LDAK. Binary phenotypes should only take values 1 (control), 2 (case) or NA. When a phenotypic values is NA for a particular sample, then that sample is excluded from the analysis; however, if a covariate is NA, its value is replaced by the mean.