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

From |
P K <statistics_2009@yahoo.de> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
st: WG: interpretation of interactions with only one variable in log format |

Date |
Thu, 4 Jun 2009 16:34:56 -0700 (PDT) |

Hi, I have two questions concerning the interpretation of an OLS regression with interaction effects, while only ONE of the interaction terms is in log format. 1) I have different dummies in my regression, and I interact each dummy with a logged continuous variable. How do I have to interpret the results: ex. MODEL 1: without interaction effects (dummies only): dummy 1 = -10 dummy 2 = -1.16 dummy 3 = 21 MODEL 2: with interaction effects: dummy 1 & log(var1) = -18.49 dummy 2 & log(var1) = -28.98 dummy 3 & log(var1) = 43.83 How do I have to interpret the results of MODEL 2, in comparison to MODEL 1? => e.g. a 10% change in dummy1*log(var1) leads to a change in the dependent variable of -18.40*log(1.1) = -0.76, which is much weaker than dummy 1 effect alone (-10) => interaction term weakens effect of dummy 1 on dependent variable? 2) I run the same models with a continuous variable instead of all dummies. MODEL 1: without interaction effects (continuous predictor variable only): independent variable = -3.0 MODEL 2: with interaction effects: independent variable * log(var1) = -4.78 How do I have to interpret the results of MODEL 2, in comparison to MODEL 1? => e.g., a 10% change in independent variable * log(var1) would lead to a change in the dependent variable of -4.78*log(1.1)=-1.978? => This does, however, not take into account that the independent variable hasn't been in log format. Should I rather use log(independent variable) for MODEL 1 and MODEL 2 and then compare the 2 models? Thanks, Pat * * 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/

- Prev by Date:
**Re: st: suggestions on a model** - Next by Date:
**st: Autocorrelation tests for FE models** - Previous by thread:
**st: Rename variables to reflect values?** - Next by thread:
**st: appling string functions across observations** - Index(es):

© Copyright 1996–2014 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |