Hi, I have a question on the calculation of linear predictor of weibull model in general (using survreg function for ex). I fit my covariate in univariate mode within cross-validation. My question is that for each covariate, I get an intercept since the covariates are fitted in univariate mode. Then, the covariates could be selected based on the smallest p-value (returned by survreg). To calculate the linear predictor of samples (or observations) in the testing set, would the sum of samples weighted by weibull coefficients be correct as linear prediction or should I somehow make use of intercepts? look forward to your reply, Carol