Fitting generalized linear model to multiple imputed datasets and perform bootstrapping
Dear all,
I am a health economist and currently working on a project to assess the cost-effectiveness a new strategy compared to usual care based on individual level data from 1020 patients.
After performing 40 times of multiple imputation, a dataset consisting of the original 1020 observations and additional 40800 observations was created, which included patient ID, treatment allocation (tr), total cost (total_cost), total QALY (total_QALY), baseline cost (cost_M0) and baseline EQ-5D utility (eq5d3l).
I have fitted two GLM to total cost and total QALY, controlling for baseline covariates:
forvalues i = 1/40 {
mi xeq `i': glm total_cost i.tr cost_M0, family(gaussian) link(identity)
mi xeq `i': glm total_QALY i.tr eq5d3l, family(gaussian) link(identity)
}
Now I am interested in the mean incremental cost and incremental QALYs between treatment arms based on 1000 times of bootstrapping. Meanwhile, the increments should also reflect the within-imputation and between-imputation variance due to multiple imputation.
I am struggling to write the code to implement the above steps, especially how to handle multiple imputation and bootstrapping . Any hints/help is highly appreciated!
Qian
I am a health economist and currently working on a project to assess the cost-effectiveness a new strategy compared to usual care based on individual level data from 1020 patients.
After performing 40 times of multiple imputation, a dataset consisting of the original 1020 observations and additional 40800 observations was created, which included patient ID, treatment allocation (tr), total cost (total_cost), total QALY (total_QALY), baseline cost (cost_M0) and baseline EQ-5D utility (eq5d3l).
I have fitted two GLM to total cost and total QALY, controlling for baseline covariates:
forvalues i = 1/40 {
mi xeq `i': glm total_cost i.tr cost_M0, family(gaussian) link(identity)
mi xeq `i': glm total_QALY i.tr eq5d3l, family(gaussian) link(identity)
}
Now I am interested in the mean incremental cost and incremental QALYs between treatment arms based on 1000 times of bootstrapping. Meanwhile, the increments should also reflect the within-imputation and between-imputation variance due to multiple imputation.
I am struggling to write the code to implement the above steps, especially how to handle multiple imputation and bootstrapping . Any hints/help is highly appreciated!
Qian
