Description of column-types used to define covariates

CONTINUOUS COVARIATE (formerly COV): continuous covariate

The column-type CONTINUOUS COVARIATE is used to tag continuous covariates, i.e covariates which can take values on a continuous scale (such as weight or age). Covariates tagged in the data set can (but don’t have to) be used as covariates in the model. The covariate value must be constant within subjects (or within occasions if occasions are present). If the value is not constant, the first value of each subject in time ordering will be used for all times (and a warning is generated). To define time varying covariate, use REGRESSORS (example here).

The allowed values are doubles or dot ‘.’. There must be at least one non-dot value per subject (or per occasion if occasions are defined).

Examples:

  • Basic example: in the following data set, individual 1 has weight WT = 78, individual 2 has WT = 80 for all times points (first value in time ordering) and individual 3 has WT = 90.
ID  TIME  Y    WT 
1    0   5.7   78   
1    1   5.6   78 
2    0   6.7   80 
2    1   6.5   82 
3    0   7.8    . 
3    1   8.9   90

Format restrictions:

  • Continuous covariate columns shall contain either a double or “.”.
  • The covariate must be defined at least once per subject-occasion.
  • The covariate must remain the same for all the lines within the same subject-occasion.

 

CATEGORICAL COVARIATE (formerly CAT): categorical covariate

The column-type CATEGORICAL COVARIATE is used to tag categorical covariates, i.e covariates which can only take a finite number of values (such as sex or a genotype). Covariates tagged in the data set can (but don’t have to) be used as covariates in the model. The covariate value must be constant within subjects (or within occasions if occasions are present). If the value is not constant, the first value of each subject in time ordering will be used for all times (and a warning is generated). To define time varying covariate, use REGRESSORS (example here).

The allowed values are strings, doubles or dot ‘.’. There must be at least one non-dot value per subject (or per occasion if occasions are defined), otherwise the category is considered to be ‘NA’.

Examples:

  • Basic example: in the following data set, the covariate SEX has two categories ‘F’ and ‘M’ and the individual 3 has value ‘F’. The covariate GENO has three categories ‘1’, ‘2’ and ‘NA’ (coming from the individual 3 which has no valid value). The covariate RACE has only one category ‘1’. Individual 1 has RACE = 1, as this is the value defined first in time. RACE will not appear in the graphical user interface because it is not possible to use a covariate with only one category in a model.
ID  TIME  Y  SEX GENO RACE
1    0   5.7  F   1     1
1    1   5.6  F   1     2
2    0   6.7  M   2     1
2    1   6.5  M   2     1
3    0   7.8  .   .     1
3    1   8.9  F   .     1

Format restrictions:

  • The covariate must be defined at least once per subject-occasion.
  • The categorical covariate must be the same for all the lines with the same subject-occasion.