Programs for Estimating Sampling Uncertainty for Cost-Effectiveness Analysis
The .zip file on this page contains .do files, datasets, and documentation for assessing sampling uncertainty that arises in cost-effectiveness analysis. The .do files contain programs that calculate confidence intervals for cost-effectiveness ratios, acceptability curves, net monetary benefits, and value of information. There are other .do files which contain programs for plotting many of these measures.
As indicated in READ ME FIRST.docx, the .do files make NO calculations themselves. “Doing” them loads the sampling uncertainty programs into memory for Stata access. For example if you type:
do inprogsz
you are loading 8 “immediate form†sampling uncertainty programs into Stata. You make calculations by typing the command lines for these programs, such as for the confidence interval for the cost-effectiveness ratio:
fiellerzi 1000 500 .1 .05 0.03 0.95
where 1000 and 500 are the difference in cost and its standard error, 0.1 and 0.05 are the difference in effect and its standard error, 0.03 is the correlation between the difference in cost and the difference in effect, and 0.95 is the width of the confidence interval that is being calculated (i.e., the 05% CI).
Many of the programs base calculations on parameters such as the difference in means, the standard error of the difference, and the correlation of the difference in means, which can be estimated by use of any number of statistical methods. For these types of programs, examples in the documentation report the parameter estimates that are being assumed in the example.
Some of the programs (in uprogs.do) use t-tests to estimate the difference in costs and effects in a dataset that is being analyzed. In the documentation, examples of these programs use the eeict11.dta dataset that is included in the .zip file. These programs will work with any dataset that has 3 variables that represents per-patient costs, per-patient effects, and a treatment group.
Other of the programs (in bsceaprogs.do) take the results of a bootstrap analysis or a second-order Monte Carlo simulation and use them to make the calculations. In the documentation, examples of these programs use the bsceaexample.dta dataset that is included in the .zip file. These programs will work with any dataset that has variables that represent replicates of the difference in cost and difference in effect (each observation represents a difference in cost and a difference in effect typically from either a bootstrap analysis or a second order Monte Carlo simulation).
For background about confidence intervals for cost-effectiveness ratios, acceptability curves, net monetary benefits, and value of information, see chapters 8 and 9 in Glick, et al., Economic Evaluation in Clinical Trials, the lectures on sampling uncertainty in the EPI 550 Medical Decision Making and Clinical Economics and Lectures by Invitation tabs, or articles on sampling uncertainty cited in the Publications tab.
THE .ZIP FILE
Back to Stata Files page