Sunday, May 5, 2013

Heteroscedasticity

On the recent class presentation, the discussion about the nature of the data, specially its variation, was brought up many times. Heteroscedasticity (pronunced heterokedastic) refers to a different kind of variation in the data, while Homescedasticity refers to a constant variation.

The non-constant variation of our data is a statistical problem that can invalidate most statistical tests of significance given the harsh effects it has when trying to execute a linear regression model. Heteroscedasticity can either bias upward or downward the results, and sometimes, its not possible to see this with a simple scatter plot.  The next time you guys have to do a regression, dont forget to account for the variation on your data; but inn case you do not have an heteroscedasticity problem, dont forget to consult your professor on how to address the problem

-Xavi

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.