After using comBat to remove know batch effects, I used sva to estimate unknown sources of variation. I saved the surrogate variables to my phenotype file and regress these variables along with other covariates as shown in the formula.

x=lm(loc2[,i]~Age+Sex+surrogate.1+surrogate.2+batch, data=pData) 

My questions are:

1.Is it appropriate to use surrogate variables in the regression model?

2.using the batch variable again since it was used in the comBat correction?