2 4 Consider the covariates diab booze recex male race smokever lowed a binary

# 2 4 consider the covariates diab booze recex male

This preview shows page 2 - 5 out of 9 pages.

2
4)Consider the covariates diab, booze, recex, male, race, smokever, lowed, (a binary variable we created in lab 6 for variable school), bmi, and age. Build an adjusted Cox model with hicholas theexposure of interest. Justify your model selection and report the adjusted estimate and 95% confidence interval for hichol. Has the estimate or interpretation changed from the unadjusted model? (Note: for the purpose of this homework, you don’t need to set the reference of age to 65).
5) Fit a Poisson survival model for the same data, with a dummy variable for each year of follow- up, using the same variables you’ve selected in question #4. How do the coefficients compare with the Cox model? /* Q5 */ proc genmod data =HW6; class fup_yr recex / param = ref ref = first; model event = hichol diab recex male smokever lowed bmi age fup_yr / type3 error =poisson link =log scale =deviance; run ; Parameter D F Estimat e Standard Error Wald 95% Confidence Limits Wald Chi- Square Pr > ChiS q Intercept 1 -11.4982 0.1023 -11.6987 -11.2977 12632.0 <.0001 hichol 1 0.0528 0.0160 0.0214 0.0841 10.90 0.0010 DIAB 1 0.6347 0.0239 0.5879 0.6815 706.21 <.0001 RECEX 2 1 0.0192 0.0249 -0.0295 0.0679 0.60 0.4398 RECEX 3 1 0.2840 0.0245 0.2360 0.3321 134.10 <.0001 MALE 1 0.4956 0.0173 0.4617 0.5294 823.22 <.0001 SMOKEVE R 1 0.5153 0.0180 0.4800 0.5505 823.01 <.0001 lowed 1 0.0882 0.0161 0.0566 0.1198 29.96 <.0001 bmi 1 -0.0169 0.0017 -0.0202 -0.0135 98.12 <.0001 age 1 0.0881 0.0009 0.0863 0.0900 8658.78 <.0001 In the Poisson model, the incidence rate ratio estimate for someone with high cholesterol is 1.0542 compared to someone with low cholesterol with a 95% CI = (1.0216, 1.0877). The coefficient here = 0.0528, whereas the Cox coefficient = 0.0562, which is similar. However, we have a significant finding for high cholesterol because the Poisson model is more efficient when the variables are correctly specified. The other covariate coefficients are roughly the same.

#### You've reached the end of your free preview.

Want to read all 9 pages?

• Spring '14
• Hernandez-Diaz
• Survival analysis, Proportional hazards models, Disadv