Hi, I am trying to use stepAIC to do forward stepwise selection for a logistic model. I did hi <- glm(melanoma ~ ., family = binomial, data=data) lo <- glm(melanoma ~ 1, family = binomial, data=data) stepAIC(lo, scope=list(upper = hi , lower = lo), direction = "forward" ) But the full model hi did not converge since I have many many candidate covariates: Warning message: algorithm did not converge in: glm.fit(x = X, y = Y, weights = weights, start = start, etastart = etastart,… Question: Can I do the forward selection without specifying a maximal model hi? Thanks, Minya Pu [[alternative HTML version deleted]]
Hi, I am trying to use stepAIC to do forward stepwise selection for a logistic model. I did hi <- glm(melanoma ~ ., family = binomial, data=data) lo <- glm(melanoma ~ 1, family = binomial, data=data) stepAIC(lo, scope=list(upper = hi , lower = lo), direction = "forward" ) But the full model hi did not converge since I have many many candidate covariates: Warning message: algorithm did not converge in: glm.fit(x = X, y = Y, weights = weights, start = start, etastart = etastart,… Question: Can I do the forward selection without specifying a maximal model hi? Thanks, Minya Pu [[alternative HTML version deleted]]