control function approach

Dear All,
Riassumo gli steps:
I have doubts on the procedure I'm using to estimate a logit fe on panel data with a continuous endogenous variable. I'm trying to follow a 'control function approach' as follows:

1. xtreg for the continuous variable subspected to be endogenous: xtreg howlng i.hsroom age agesq i.nkids i.nwage, fe, if sex==2
2. calculation of the residuals: predict ue
3. inclusion of the residuals in the logit fe (without the variable subspected to be endogenous, howlng, and its instruments used in step 1), i.hsroom age agesq i.nkids i.nwage: xtlogit in_work i.numkids i.qfachi i.wave i.spjb i.grp ue howlng_part if sex==2 & cpstat==3, or

Do you this am I correct? the results make sense apart the standard errors that for some covariates are very high, and for this reason I'm asking whether this is the correct procedure.
Best,
Chiara