I have a case-control study with 200 cases and 200 controls (d == 0 for controls and d == 1 for cases). Two tests y1 and y2 are applied to each case and control and the results are coded 1 for a positive test result and 0 for a negative test result. The estimated prevalence of the disease in the population, p = 0.15.
I want to calculate the PPV and NPV of these tests, the 95% CI of the predictive values, and compare the PPV and NPV of tests y1 and y2.
Can someone please help me with some Stata code for the same?
I am attaching the data file for the exercise. This is hypothetical data generated for an academic exercise only.
Hello Inaamul Haq. The resource that comes to mind is Robert Newcombe's book:
- https://www.routledge.com/Confidence-Intervals-for-Proportions-and-Related-Measures-of-Effect-Size/Newcombe/p/book/9780367576707
Excel workbooks to carry out the calculations can be obtained here:
- http://profrobertnewcomberesources.yolasite.com/
I don't see a workbook for comparing PV+ and PV- (or PPV and NPV as you called them) for paired tests; but there is a workbook for comparing Sens and Spec from paired tests. The reference shown for that workbook is:
Statistics in Medicine 2001, 20, 907-915
If you look up that article, you might be able to work out how to tweak things to get the comparisons you want. (And if it can be implemented in Excel, it can be programmed in Stata!)
My other suggestion is that for questions like, "How can I do X in Stata?", Statalist is arguably a better place to post your question. If you do decide to post there, please review the FAQ first, as it has some great advice on how to make your question more answerable.
- https://www.statalist.org/forums/forum/general-stata-discussion/general
- https://www.statalist.org/forums/help
Narjis Hadi Irhaif, you need to provide far more detail about how to "access the appropriate analysis" in SPSS. I am a fairly experienced SPSS user, and I know of no commands that will compare the PV+ and PV- values from two diagnostic tests administered to the same people. If one is required to find the equations somewhere and write code to carry out the comparison, that can be done just as easily using Stata. Thank you for clarifying your advice and providing the needed detail.
Thank you Bruce Weaver . Robert Newcombe's book looks very useful. Thanks for the link.
In a case-control study direct calculation of PV+ and PV- cannot be made; disease prevalence must be known to estimate PV+ and PV-. The diagt command in Stata (install using ssc install diagt) estimates PV+ and PV- and their 95% CI in a case-control design. The option prev(#) enables this estimation. However, I could not find any direct command to compare the PV+ and PV-.
The Excel Workbook links provided by Bruce Weaver are helpful and can solve this issue.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis