Missing values in explanatory variables


ASReml will abort the analysis if it finds missing values in the design matrix unless !MVINCLUDE or !MVREMOVE is specified,
     !MVINCLUDE causes the missing value to be treated as a zero.
     !MVREMOVE causes ASReml to discard the whole record.


    1 Identify which model terms have the missing values from the data summary,
    2 Understand why observations are missing,
    3 Decide whether to delete the records, centre covariables, impute a value to the missing observations or treat missing treatment levels as an extra category.

Important: Use !MVREMOVE if you are not sure that !MVINCLUDE is appropriate.

While !MVREMOVE is an easy way to discard records with missing explanatory variables, if it causes the job to fail for other reasons, it may be necessary to explicitly drop records containing missing values in particular fields with the !D transformation.

The !NA transformation can change missing values to a nominated value.


Treating missing values as in covariates is usually only sensible if the covariate is (has mean of zero).

Design factors

the factor level is zero (or missing and the !MVINCLUDE qualifier is specified), no level is assigned to the for that record.

Return to start