I am using the Big Class example for my demonstration. I clear row states so that all rows have their default row state. I launch Bivariate with weight (Y) versus height (X) through Analyze > Fit Y by X. This platform does not respond to Exclude row state changes by default so click the red triangle and select Redo > Automatic Recalc. I click the red triangle and select Fit Line. The results use and show all the observations (rows).

I select and hide the observations for females.

The hidden rows disappear from the scatter plot but the regression remains unchanged.
I unhide the selected rows and then exclude them.

All of the data is shown but the regression uses only the 22 observations for males.
Is this kind of control not what you want?