how does PROC SURVEYREG (and other regression procedures) treat observations with missing value?

[somebody]

So say some of the variables have missing values beause (i)the value is missing, and (ii) the value is 0. Does SAS skip these observations? 

[Reeza]

In the majority of procedures, any missing value in the row, within  will cause it to be excluded entirely.

[ballardw · Accepted solution]

Variables referenced in the model statement, by, class, weight and in the case of the survey procs strata or cluster that are missing will have the record excluded from the analysis. In some cases, class variables for example, can be set to treat missing as a category (though all of the missing will be treated as a single group value).

A value of zero will be treated as a valid value unless on a weight statement (generally). If you are using zero for "missing" where it does not have a meaning of zero associated with the other values for that variable (height of a person comes to mind where zero is unlikely) then you might want to reconsider the value, at least in modeling proc.

[Reeza]

Note that in addition to @ballardw statement, if the variable is listed in the CLASS statement but not actually used in the model it still is included in the list of missing that will cause a row to be excluded.