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

From |
Maarten buis <[email protected]> |

To |
[email protected] |

Subject |
Re: st: Significance of Probit estimates versus marginal effects |

Date |
Thu, 23 Oct 2008 08:45:31 +0100 (BST) |

--- Kir1 <[email protected]> wrote: > When i run a probit I find the variable of interest statistically > significant. However, once I list the marginal effects using > 'margeff', the variable is not statistically significant anymore. > (margeff, at(mean) same as mfx is still significant as the probit > estimate, albeit at a different level of significance). > > What can i make of this?Is the effect of the variable significant or > not? what is the different meaning of the significance of the probit > estimate versus that of the marginal effect? These two tests test different hypotheses: The coefficient in a probit model tells you the effect of a variable on the latent propencity for a positive result, while while -margeff- and -mfx- will give you an effect on the probability of a positive outcome. The effect on the probability is usually the thing of interest, however the size of this effect (and thus whether or not it is significant) depends on at what values of the explanatory variables you evaluate that effect. In other words each observation has it's own effect on the probability, which depends on the values of all its explanatory variables. One of the advantages of -margeff- over -mfx- is that -margeff- allows you to compute the average effect rather than the effect at average values of the explanatory variables. The distinction here is between the effect for a typical individual and the effect of someone with typical values on it's explanatory variables. However, because you have specified the -at(mean)- option you are using the same concept of effect as -mfx-. So, what should you do? Choose which effect you are interested in, the effect on the latent propensity (-probit-), the effect on the probability for someone with typical values on the explanatory variables (-mfx- or -margeff- with the -at(mean)- option), or the effect on the probability for a typical person (-margeff- without the -at(mean)- option). Choose your null hypothesis, typically that would be that the effect of interest equals zero. Then report whether or not you reject that hypothesis. (When we are saying that an effect is significant, we are implicitly saying we went through all these steps and concluded that we rejected the null hypothesis) Hope this helps, Maarten ----------------------------------------- Maarten L. Buis Department of Social Research Methodology Vrije Universiteit Amsterdam Boelelaan 1081 1081 HV Amsterdam The Netherlands visiting address: Buitenveldertselaan 3 (Metropolitan), room N515 +31 20 5986715 http://home.fsw.vu.nl/m.buis/ ----------------------------------------- * * 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**:

- Prev by Date:
**st: RE: doubt how to avoid some problems with do files** - Next by Date:
**RE: st: mfx after xtnbreg and how to compute predicted Y** - Previous by thread:
**st: Significance of Probit estimates versus marginal effects** - Next by thread:
**Re: st: Significance of Probit estimates versus marginal effects** - Index(es):

© Copyright 1996–2024 StataCorp LLC | Terms of use | Privacy | Contact us | What's new | Site index |