Marcello and Maarten,
Thanks for your suggestions and comments. I have to do some thinking,
since I thought that my customer wanted a graph of the probability of an
"average person living in year X" by year. I had not foreseen my
technical question would invoke a statistical debate.
Question:
* Why do all procedures developed to produce effect displays provide the
opportunity to fix covariates at their means (or a factors at its
proportional distibution in the data), e.g. postgr3 in STATA, effects in R?
Another thought on a seemingly unrelated topic (just to spice up the
discussion):
* In the multiple imputation literature, it is suggested that the
imputation procedure should allow binary covariates to be imputed with
values ranging between 0 and 1. This means that missings on e.g. gender
might indeed be imputed as 0.79 or 0.13. Apparently, we should think of
these people as half-man, half-woman.
Stijn
Stijn Ruiter
Department of Sociology
Radboud University Nijmegen
P.O. Box 9104
6500 HE Nijmegen
Netherlands
Phone: + 31 24 361 2272
Fax: + 31 24 361 2399
Visiting address:
Thomas van Aquinostraat 4.01.74
Nijmegen
website: http://oase.uci.ru.nl/~sruiter
Marcello Pagano wrote:
>Flavours aside, Maarten, I disagree that this gives the average over
>cohorts. It is not an average. It is the probability of an `average' person.
>
>a) This is exactly my point. Taking average heights yields the average
>of a physical property. That no one exists who has that height is not
>important. If, on the other hand, you wish to construct such a person,
>for whatever reason, and I cannot think of one, fine. It is not
>necessary since you have the average height. Life, of course, gets silly
>when we think of a person with average binary covariates: half-dead, etc...
>
>b) Why it would "be a better description of the typical predicted
>probability" goes back to why an average is a good predictor of a
>variable. Same context. You want a "typical" probability. Each person
>has a probability fitted to her/him. This is just like any other
>characteristic (height, weight etc...). Take the average of these, like
>any good statistician, to describe the typical, however you take averages.
>
>m.p.
>
>
>
>Maarten Buis wrote:
>
>
>>Marcello:
>>A reasonable argument can be made for Stijn's position, if the
>>mean changes over cohorts, e.g. the proportion of mothers that
>>are working. It would show the change over cohorts, including
>>the change in the distribution of working status of the
>>mothers. In this sense this approach has a clear "population
>>average flavour".
>>
>>There are however clearly some issues with this approach:
>>a) It is true that the person with average values on the
>>explanatory variables cannot exist, but we almost never think
>>the "average person" is a real individual. This is just a
>>construct that helps us summarize what we see.
>>
>>b) It is true that the predicted probability for an
>>individual with average values on the explanatory variables is
>>different from the average predicted probability, but I don't
>>see why one would be a better description of the typical
>>predicted probability than the other.
>>
>>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 Z434
>>
>>+31 20 5986715
>>
>>http://home.fsw.vu.nl/m.buis/
>>-----------------------------------------
>>
>>
>>
>*
>* 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/
>
>
>
*
* 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/