Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Holly E Reed <Holly.Reed@qc.cuny.edu> |
To | "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |
Subject | Re: st: Re: How do I create a calendar year variable by person id before reshaping to person-year dataset? |
Date | Thu, 13 Feb 2014 15:29:09 +0000 |
Well, thank you for trying, Nick. I appreciate you following up. I'll let Statalist know if I ever solve the problem. Best regards,Holly __________________ For completeness: sorry that I didn't reply, I didn't see that your extra detail changed my impression that your data structure is fine as it is. I was hoping that someone who does more of what you do would chime in, but they didn't. I couldn't comment on your failed attempt to replicate what you did previously. Nick njcoxstata@gmail.com On 12 February 2014 19:08, Holly E Reed <Holly.Reed@qc.cuny.edu> wrote: > I am reposting this as I never received a reply to my last question. I hope that someone can help me to figure out the best way to approach this. > > I am trying to create a person-year dataset for use in a binomial logit event history model. I want to have one observation for each year of a person's life (one record per person-year), but I also need the calendar years as the variables, so that I can create covariates such as whether a woman had a birth in a particular year (0/1 dummy), whether a woman moved in a particular year (0/1 dummy), etc. There will be some non-time-varying covariates, of course, but for the data where I do have time covariance, I want to create dummies for events in each specific year.I will probably create parity as a covariate also, as Steve mentions, but the main question of interest is whether a woman gives birth in a particular year (this will be lagged by one year) and whether that predicts the probability of migration in the next year. > > Does that make sense? If you have a better suggestion, please let me know. I did create a migration dataset from this same data several years ago, but I did not have the birth data at the time and I simply want to merge it with the migration dataset I already have, but I need it to be in the same format and I can't seem to get it work the way it did the last time. > > Thanks, > Holly > ______________ > As you point out, Nick, Holly has not told us what she wants to do. I'm > not sure either, as she's tried reshapes both long and wide. > > I've analyzed many data sets of reproductive histories, and I always use > the long format. With this format, it's easy to create running totals of > different kinds, for example parity, the number of births + stillbirths > at a given time. We also can merge in dates of different risk factor > exposures and so easily assign a prior or recent exposure to each > pregnancy. In fact, we always *collect* the data in long format, > as it shortens the codebook and greatly simplifies the edit process. It > also allows a woman to recall a pregnancy out of order, because one > can re-order by date. > > Steve > sjsamuels@gmail.com > > > On Feb 6, 2014, at 6:14 PM, Nick Cox <njcoxstata@gmail.com> wrote: > > Thanks. That helps. I have already explained why -reshape wide- > doesn't work. -reshape- maps one variable to several, but the "birth" > you feed to it is the stub for several variables. > > Otherwise, I don't see why you want or need to -reshape- at all. > Several variables are repeated for each woman and some vary. Depending > on what you want to do, you either reduce the dataset by removing > duplicates or keep the whole dataset. > > Alternatively, if you explain why you (think you) need to -reshape- > that might illuminate what is being misunderstood, or what you need to > do. > Nick > njcoxstata@gmail.com <snip> * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/