Introduction Effects of sex, age, medication use etc., may be incorporated in the
means model. Without covariates the model for the means is simply a grand mean:

It is possibly in Mx to incorporate any model that one desires for the observed scores into the means model. The model specified for the variance/covariance will be evaluated simultaneously on the residuals.

For the observed score y for individual I, a linear regression of age on the and a deviation for males can be modeled ad follows;

where denotes
the grand mean,
denotes the regression weight of age,
denotes the deviation of males (if sex is coded 0 for females, 1 for males).
Observed variables that are not analysed as dependent variables, are called
'definition variables' in Mx. Note that in all Mx scripts incorporating sex as covariate in the
means model we assume sex is coded 0 and 1

Data preparation for Mx Mx expects one missing value, read as a string, for
all variables. When conducting multivariate analyses missing values are handles
efficiently, such that a missing value in one dependent variable does not
result in the loss of a non-missing second dependent variable for that person
(see also Mx manual). However, when variables that
are defined as 'definition variables', missing values in them will lead to the
loss of all the data from the whole case. Therefore, variables used as definition
variables need to have a separate missing value, other than the missing value
assigned to regular dependent variables. There is no option in Mx to define
a missing value for definition variables; the way Mx works is: it reads the
data for a case. If the dependent variable has a missing value (e.g. -99.99),
it will still read the values for the definition variables. These values may
either be existing values or may also be missing (but different value than -99.99,
e.g. -88.88). Since the dependent variable is missing, the definition variables
will be read, but not be used for the analysis.

This has several consequences:

since unused definition variables are read by Mx, the descriptives of definition
variables in Mx are non-informative

if the dependent variable is NOT missing, the definition variables MUST
also NOT be missing (i.e. they must be measured, and cannot have value -88.88
in this example)

The other way around is no problem: if the dependent variable is missing,
definition variables may be measured as well as missing.

The expected means apply to the last case read in that datagroup. If one
or two members of this case have a missing vale on the dependent variable,
the expected mean will be non-informative for this individual