The CMS models for choosing comorbidities to include in a risk adjustment model for a quality measure are based on bootstrapping the regression 1000 times and choosing comorbidities (coded as clinical conditions with 0/1 coding of the presence of the condition for each patient, there are approximately 100 clinical conditions) for the final risk adjustment model if the p-value for the estimated coefficient on the comorbidity is statistically significant 90% of the time. Regardless of the wisdom of this strategy, I need to duplicate this methodology in data I have. In order to duplicate it, I need to set a seed, bootstrap the relevant regression, which will be either a logit for 0/1 outcome, or glm with log transform for cost outcomes, and recover the p-values for each covariate in the model for each iteration, so I can count if it is <0.05 in 90% of the cases. I am not sure how to bootstrap with replacement and save the p-values for each covariate from each iteration in a file or matrix so I can examine them. Any suggestions on how to code this would be appreciated.