Timeline for In a census-like situation, does it make sense to use Bayes Rule?
Current License: CC BY-SA 3.0
10 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| May 8, 2012 at 21:26 | vote | accept | Wayne | ||
| May 8, 2012 at 20:42 | history | edited | Wayne | CC BY-SA 3.0 |
added 160 characters in body
|
| May 8, 2012 at 16:41 | comment | added | Wayne | @Aniko: I've edited my question, but when I wrote it I was thinking of a simple, textbook-like case: a single binary question on the survey. You have a prior survey's results and the new survey's results, yielding a posterior. Does that thinking make sense. (Several answers make it clear, it doesn't.) | |
| May 8, 2012 at 16:36 | history | edited | Wayne | CC BY-SA 3.0 |
added 1141 characters in body
|
| May 8, 2012 at 16:01 | answer | added | ely | timeline score: 2 | |
| May 8, 2012 at 15:27 | comment | added | Michael R. Chernick | It would be used to impute. Then a Bayesian approach along the lines of what I surmised could be used to characterize the uncertainty in the second year sentence due to the imputation. This would also depend on the method of imputation. I think the simple rule of LOCF might not be the best as you may be able to borrow strength fromthose that responded in years 2. | |
| May 8, 2012 at 15:21 | comment | added | Michael R. Chernick | @Aniko My guess is that it the scenario he uses one years results to get a prior uses the second years results for the likelihood and "runs the results through Bayes rule" to get a posterior distribution. But what inference problem is the posterior there for. The question is a bit too vague. I do think that if a complete census is obtained both times then you could look at how answers changed on individual and group responses. But there is no random sample and no inference problem. If data is complete in the first year but not in the second the information from the first could be used. | |
| May 8, 2012 at 15:20 | answer | added | Erik | timeline score: 3 | |
| May 8, 2012 at 13:42 | comment | added | Aniko | What do you mean by "running the results through Bayes rule"? What question are you trying to answer? | |
| May 8, 2012 at 13:37 | history | asked | Wayne | CC BY-SA 3.0 |