Hi Rich, Thanks for your interest. "Rich Ulrich" <[EMAIL PROTECTED]> wrote in message [EMAIL PROTECTED]">news:[EMAIL PROTECTED]... > On Wed, 15 Mar 2000 22:48:01 GMT, "JP" <R E M O V E > [EMAIL PROTECTED]> wrote: > ... > > I am in the process of analyzing data of such type: > > We have data on 1800 doctors over 49 months:few dependent variables (a > > particular drug prescription level), few independent (some time related > > (severity of patients seen in the months, practice volume,...) and some > > constant over time: university, sex, and years of practive (which can also > > be considered as time dependant)). > > > > id month y timeind1 ... timedep1 ... > > 1 1 4 1 30 > > 1 2 6 1 36 > < snip, lines of eample > > > Basically, a bulletin was introduced at month 37, we want to assess if this > > bulletin had an effect on a particular drug prescription pattern (y). > > What we plan to do is to model y in terms of the dependant variables based > > on the first 36 months, and then forecast (including a CI) on the last 13 > > Huh? Possibly, I am being too hostile to time-series analyses, but it > looks to me like there ought to be a simpler approach. > Another group did a similar study (in design). They used Cross-Sectionnal Pooled Time Series Regression (sounds like a full course menu, isn't it). It is not common. I never heard of it before doing this project. The other group studied the effect of the new Quebec insurance plan. We have the effect of the plan in your data too. It seems to create more variablity than the bulletin that we are studying. As you believed, it is more hopefully more related to multiple regression than to proper time series analysis. The term time series is mentionned because this technique allows to study the evolution of a dependant variable accross time, taking into account some individual characteristics. And also because you can include AR() terms in the model to account for time dependancy. > What do you know already? What do you expect? The prescribing patterns (depending on doctors' gender, year of graduation, university, patients' characteristics: age, gender, ...) are quite known. We can do for exemple some multivariate regression for only 1 given time to relate drug prescribing to mds and patients' characteristics. But that is already quite known. We are more interested in the time trends, depending on doctors' characteristics. The first goal is to assess if the bulletin had an impact: So we will build a model explaining the level (and hence the evolution as it is explained accross time) of the Y (say # of prescriptions of Benzodiazepines) by the individual characteristics (+some time effects: dummy vars for seasonality for exemple). the second goal is to look for diffrent trends according to mds characteristics. Regading the expectations,... For exemple we expect younger mds not to lower the prescribing of the studied drug, because they have younger patients and they are better educated related side effects of the drug. Older mds have patients they've treated for a long time, and are less likely to change their patients prescriptions, if the patients are already stable. So I am looking for someone who has some experience with such analysis, and could show me the different pitfalls I should be aware of. JP > That is, how much pattern is there in the early 36 months; and, can't > that be reduced to an interesting variable or two.... > > Siimilarly, what patterns do you expect in the 13 months after; and, > can't *those* be reduced to a variable or two.... > > Then, you have just a handful of variables for each doctor, and 1800 > doctors. That should give you quite a lot of ability to describe your > interesting patterns, and whatever relates them, using various > univariate and multivariate tools. > > IF you really want the time-series analysis, it sounds like you need > to bring in someone with experience at it. > > Also, for your other question, there is > http://www.stattransfer.com/lists.html > > which has information about joining various mailing lists. > > -- > Rich Ulrich, [EMAIL PROTECTED] > http://www.pitt.edu/~wpilib/index.html =========================================================================== This list is open to everyone. Occasionally, less thoughtful people send inappropriate messages. Please DO NOT COMPLAIN TO THE POSTMASTER about these messages because the postmaster has no way of controlling them, and excessive complaints will result in termination of the list. For information about this list, including information about the problem of inappropriate messages and information about how to unsubscribe, please see the web page at http://jse.stat.ncsu.edu/ ===========================================================================
