You have not said what software you are using. But in general, you can always run a regression with predictors "predfev", "gender", and their interaction. That regression will give you the appropriate coefficients to draw two lines, and then you can draw the two lines in whatever way you would normally draw lines in this software.

This is on STATA 13

for a linear model,  you can fit for predictors and their interactions, usually by explicitly providing the terms in the lmfit equation (in R).

For stat, see this:

https://statistics.laerd.com/stata-tutorials/multiple-regression-using-stata.php

you can change the 'main effect' to to include interactions.

alternately, you can explicitly create the interaction terms yourself before regression

also, see this

http://www.stata.com/meeting/germany12/abstracts/desug12_royston.pdf

Got it.  Thanks!!

On stata you can do this with the following command.

twoway ((scatter fvc predfvc if Sex ==0,msymbol(Oh))(lfit fvc predfvc if Sex==0)(scatter fvc predfvc if Sex ==1,msymbol(S)) (lfit fvc predf

> vc if Sex==1))

consider fvc and predfev as the variables X and Y you want to subgroup by categorical variable 'sex'..