Bruce Weaver wrote: > Alan Acock wrote: > >> SPSS does have the most friendly menu system, but it is a long way from >> appropriate to public health. Art Kendall does not mention Stata that >> also >> has a nice menu system but has vastly more capabilities for public >> health in >> terms of epidemiological methods and limited dependent variables. SAS has >> the steepest learning curve, but also has the most overall capabilities. >> SPSS is also broken into a series of modules and unless your >> university can >> afford all of the modules you are missing major components such as >> missing >> values or complex surveys. >> >> Version 13 of SPSS may add more capability to its complex survey >> capabilities but if you want to do more than a cross tab or estimate >> means, >> and you have samples that are not strictly random, then SPSS is simply >> inappropriate. >> >> The syntax of SPSS is hardly easy to read and the syntax generated by the >> menu system includes a huge amount of unnecessary but confusing code. >> Also, >> the various procedures use different syntax reflecting many authors >> who were >> uncoordinated. If you've seen syntax for SPSS, you should look at a >> modern >> program such as Stata. To do multiple OLS regression of y on x1 x2 x3 >> and x4 >> you would enter regress y x1 x2 x3 x4. To do a logistic regression you >> would >> enter logit y x1 x2 x3 x4. To do a multinomial logistic regression you >> would >> enter mlogit y x1 x2 x3 x4. To do a zero inflated Poisson regresion you >> would enter zip y x1 x2 x3 x4. ETC.! >> >> SPSS is a static program that is revised every couple years with most >> revisions going into gee wiz graphs for marketing. It lags behind several >> competitors in integrating new statistical procedures. Stata is updated >> monthly (command: update all) and has many powerful user contributed >> procedures. >> Alan Acock > > > > Hi Alan. Any comments on Stata's data manipulation capabilities? > > Cheers, > Bruce
I've been playing back and forth between SAS and Stata with respect to data manipulation. Let me start with this, if you have datasets on the order of several gigabytes, Stata will likely choke and die without alot of extra programming tricks. This is because Stata loads the entire data matrix in to memory. That being said, if you can load the matrix into memory without resorting to a lot of disk caching/swaping, Stata hands down wins. It took me a while to come this conclusion, only because one approaches data management very differently between the packages and I was thinking like a SAS user when programming Stata. Once I began to think like a Stata user when using Stata the programming can much easier. I tend to require a whole lot less key strokes to do the same things in Stata as I do with SAS. The speed is phenomenal, the price is right (perpetual license) and the customer support is without equal. . . ================================================================ This list will soon be replaced by the new list EDSTAT-L at Penn State. Please subscribe to the new list using the web interface at http://lists.psu.edu/archives/edstat-l.html. ================================================================
