st0399_1: Estimation of mean health care costs and incremental cost-effectiveness ratios with possibly censored data. S. Chen, J. Rolfes, and H. Zhao. Stata Journal 15: 698–711.
This update to the hcost command includes the following changes:
1. The running speed of the previous version of hcost is slow when there are many subjects. The program has been optimized, and it now substantially speeds up.
2. Previously, hcost may have produced an error message of “subscript invalid” if l() was very large. This has been fixed.
3. Additional data checking prior to analysis has been added to detect potential data errors, such as inconsistent survival time and death indicators within the same subject, the same subject ID appearing in different groups, and stop time for costs occurring after death time.
4. Previously, hcost prorated costs by L-truncated follow-up time only when start and stop time of costs were provided and stop time was larger than L. This updated version also prorates costs when the total costs are provided for each subject without start and stop times by assuming that costs are evenly distributed from 0 to the follow-up time.
5. Remarks have been added to the help file about the selection of the cost truncation limit L.
st0526_1: cvcrand and cptest: Commands for efficient design and analysis of cluster randomized trials using constrained randomization and permutation tests. J. A. Gallis, F. Li, H. Yu, and E. L. Turner. Stata Journal 18: 357–378.
The following updates have been made to cvcrand since the last time it was uploaded. First, we have removed the command’s dependence on the communitycontributed table1 package (Clayton 2013). Now, users can make their own summary tables using a command of their choice, such as summtab (Gallis 2019). Second, the constrained dataset is now automatically saved so that users cannot opt out of creating this dataset. This functionality was added because for clustered permutation test analysis, it is crucial that users have the constrained space available in dataset form. Finally, we have added a check of randomization validity (Bailey and Rowley 1987). When one constrains the randomization space, certain pairs of clusters may always (or usually) be allocated to the same arm or never (or rarely) be allocated to the same arm. Either scenario can lead to loss of randomization validity. The command cvcrand now provides with specification of the validitycheck option the summary statistics for the proportion of times in the constrained space that clusters are paired together and the proportion of times that they are not paired together to help users assess randomization validity.
st0574_1: gidm: A command for generalized inflated discrete models. Y. Xia, Y. Zhou, and T. Cai. Stata Journal 19: 698–718.
The gidm command has been updated with a change to the missing option of ml, which fixes possible convergence issues.
