Re: st: Log Transform Justification

[Phil Schumm <pschumm@uchicago.edu>]

On Aug 27, 2007, at 3:43 AM, Md Bilal Hossain wrote:
> In my PhD project, I was having problem to have a good model with  
> Thikonov Regularization in inverse problem. I checked the data and  
> try to get a good distribution of the data set.  I did log  
> transform of raw data and run the model with transformed data. The  
> model that I get is fantastic. Now I am happy to put the model,  
> however, I think I have to justify the data transformation. So how  
> can I justify?

First, understand that you have not provided enough information for  
anyone to give a specific answer to your question.  The most general  
answer, of course, is that the justification is this: this is the  
model that fits the data.  Models that clearly don't fit the data are  
not useful (at least from a data-analytic perspective).

In my experience, reviewers/readers who are bothered by a  
transformation of the data come from two main groups.  The first are  
those who are having difficulty interpreting the results of the model  
(this is by far the largest group).  For those, translating your  
results back onto the original scale is usually sufficient (though be  
careful how you do this).  The second are people who genuinely have a  
problem with the model because it appears to violate some theoretical  
expectation (these are typically individuals from a mathematical  
discipline such as physics or economics).  In this case, you may be  
obligated to demonstrate that the model on the untransformed data  
does not fit well, either graphically or via a formal test.  At that  
point, one of two things must be true: either the theory is wrong, or  
your data are bad.  Either way, this gets you off the topic of your  
analysis and into other questions.


-- Phil

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/