Linked-In Q&A: 2SLS Advice

More Linked-In Q&A:

Can anyone give me guidance on performing a two-stage least-squares analysis using Excel, Minitab 15, or Statistica 9?

I’m taking an economics course and am required to perform this analysis, but only have experience in Excel. I’m not certain that Excel can handle this, so I downloaded trial versions of Minitab 15 and Statistica 9 but don’t have much experience with either program. Any guidance or instructions would certainly help. Thanks in advance…

My answer:

First, I will try to summarize what two-stage least squares regression (2SLS) is, for common understanding. 2SLS is a method for regression on models that violate one of the assumptions of standard, ordinary least squares (OLS) regression: recursivity. A path diagram (or equivalent structural equation/simultaneous equation/path equation) is said to be recursive when the causal arrows and error terms are non-looping. Basically, when there are loops in the model, an instrumental variable must be interposed to allow for the application of OLS. The steps are as follows: 1) new dependent or endogenous variables are created to substitute for the original ones, and 2) the regression is computed using OLS, but using the newly created variables, thus circumventing the resursivity constraint. For details on a concrete example, see the first link below.

By understanding the above in detail, 2SLS can be implemented in any modern-day statistical software (R, Statistica, SPSS), and even Excel with enough programming. So, what to do? Minitab does not appear to have a native 2SLS module, though you should be able to piece together a solution from its capabilities, but that is far from efficient. Apparently, Statistica 9 has an add-on module for handling 2SLS, but I have never used it. I suggest trying to get the 2SLS add-on for Statistica, if possible, or simply download EasyReg (second link below). EasyReg is free for non -commercial use, and has online instructions for ordinary least squares regression (last link below).


Leave a comment

Your email address will not be published. Required fields are marked *