[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
Maarten buis <maartenbuis@yahoo.co.uk> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: Interpreting Fairlie decomposition results |

Date |
Tue, 20 Oct 2009 12:58:45 -0700 (PDT) |

--- On Tue, 20/10/09, Stephan F Gohmann wrote: > I have used the Fairlie do file to > decompose a logit model and I am not quite sure about how to > interpret the signs on the total gap explained. If the gap > has a value of say -0.130 and the total explained is 0.017, > then should I interpret that as -13.24% of the gap is > explained by the model. Likewise, if the gap is -.130 and > the total explained is -0.092, then is the portion explained > 70.23%. So overall, if the model fits well and the gap is > negative, then the total explained should also be negative > and when the gap is positive, then the total explained > should be positive? And if the signs are opposite, then the > model is not explaining the gap. Is this the correct > interpretation? Thanks for your help. The concept of proportion of the total effect explained does more harm than good. It is better to think in terms of direct, indirect, and total effects. There is a variable X that directly influences another variable Y, but it also influences another variable Z, which in turn also influences Y. So, X has a direct effect on Y and an indirect effect, through Z, on Y. The total effect is the sum of the direct and indirect effect. There is no reason why the direct and indirect effects should have the same sign, and opposite signs in no way indicate that the model is in some way bad or lacks explaining power, it just means that the total effect would have been bigger if the indirect effect did not exist. Hope this helps, Maarten -------------------------- Maarten L. Buis Institut fuer Soziologie Universitaet Tuebingen Wilhelmstrasse 36 72074 Tuebingen Germany http://www.maartenbuis.nl -------------------------- * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

**References**:**st: Interpreting Fairlie decomposition results***From:*Stephan F Gohmann <steve.gohmann@louisville.edu>

- Prev by Date:
**RE: st: Elements of Statistical Learning** - Next by Date:
**st: merging values of one variable to to the nearest value in another dataset (stata v10.1)** - Previous by thread:
**st: Interpreting Fairlie decomposition results** - Next by thread:
**st: Post-estimation commands for Multiple Imputation** - Index(es):

© Copyright 1996–2015 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |