Heavy Lifting - thoughts and web finds by an economist
     I also contribute to Division of Labour Load HL's Front Page
Monday, June 27, 2005

My "screed" to STATA

Here is the text of an email I sent to the good folks at STATA. It seems that STATA small, which is their "student version" can't estimate a stochastic frontier model on 25 observations!! In my mind this is a level of crippling that borders on the ludicrous. Of what use is the package as an instructional tool if it can't estimate the models presented in class?

We'll see if I get any response:

I am teaching a graduate econometrics course (panel data, stochastic frontiers, maximum likelihood, limited dependent variables) and my students are having considerable problems using Small STATA for their homework assignments. I am running STATA SE, which of course precludes a lot of the size/memory issues.

My first experience with the limitations of small STATA was on a panel data problem using 50 cross sections and 10 observations each, for a total sampleof 500 observations. Small STATA was able to run fixed and random effects models but was not able to estimate XTGLS or XTPCSE models - I understand that the matrix size was too big. For this size data set, I suppose the limitations are reasonable.

However, their current homework is to estimate the stochastic frontier model presented as an example in the STATA manual using the 25 observation (!!) data sample from Greene (1990). Using only twenty five observations in a cross section, small STATA is unable to estimate the frontier model with only two right hand side variables!!!

I understand that small STATA was designed for student applications, however it is really only good for OLS and the other standard problems addressed in an introductory/undergraduate course, it is not capable of performing even example versions of more advanced estimators. Of what use is small STATA beyond OLS? Is there any way to expand the capacity of small STATA so that it can at least estimate most of the estimators on at least a small sample of data?

I am a dedicated STATA user and wish to continue using STATA in my classes but if the computer package is unable to estimate the models discussed in class, I trust you can see where it usefulness is very much limited.

Your help or consideration in this matter is greatly appreciated.


Craig Depken

Update: Why STATA rocks!!! The folks at STATA responded this afternoon and it looks like it is my fault after all...I have been extolling the virtues of STATA - especially STATA S/E - and I will stand by my claims. STATA is the best overall econometrics package out there. Some packages are a bit easier to use for certain things, but overall once you go STATA you won't go back...this coming from a recovering TSPer.

Comments: Post a Comment


Le Chai - galerie du vin



Posts that contain Craig Depken per day for the last 90 days.


Heavy Lifting's Main Page
Email Me
Atom Feed

Heavy Lifting

Great Links

Money I Found Today

Heavy Lifting - Firehose style (56k warning)

Recent Posts

- My "screed" to STATA



Site Meter Blogroll Me!


Modified maystar design
powered by blogger