Re: st: Control for the dependency of observations when correlate & ANOVA

[David Airey <david.airey@Vanderbilt.Edu>]

.

You mentioned two variables, parental education and child gender. And  
then you also mentioned correlation. ANOVA is for continuous dependent  
variables, and the typical output will not give you correlation  
directly. Maybe you just meant association between parental education  
and child gender. That's a strange but interesting association! Since  
child gender is randomly determined at conception (at least at the  
population level), your hypothesis is that child gender causes  
variation in parental education? Causation the other way around  
assumes the parents had somehow selected the sex of their children by  
other means. And most parents are done with there education before  
they give birth, so why do you expect an association? Do you have  
enough variation in your education variable to treat it as a  
continuous variable? Even if not, you can find a model that associates  
parental education (continuous, ordinal, nomimal) with child gender  
(nominal binary) with the cluster option in Stata. I think you could  
also find a mixed model in Stata too (xtmixed, xtmelogit, gllamm).

-Dave

On Jan 9, 2008, at 8:40 AM, Wen Jun wrote:


> Thanks. It's another learning. :) - wj
>
>
>
> On 1/8/08, David Airey <david.airey@vanderbilt.edu> wrote:
> > .
> >
> > Oops. Sorry. I read your post too quickly and even inserted the fact
> > that you might be studying twin data or genetics when you did not say
> > that!
> >
> > -Dave
> >
> > On Jan 8, 2008, at 5:48 PM, Wen Jun wrote:
> >
> > > Hi All,
> > >
> > > I am trying to run some descriptive analyses (i.e. correlations,
> > > ANOVA) on two children within one family and want to control for the
> > > dependency of observations.  For example, my data include 1000
> > > families and each family has two children. I want to know the
> > > correlation between parenatl education and child gender, but want to
> > > control for the dependency of observation. That is, I don't want to
> > > correlate parental education and child gender for first and second
> > > born specifically. I am a SAS user. I have heard that this is  
> > > possible
> > > to do in STATA, although I have no idea how to do it. Please give  
> > > some
> > > suggestions. Thanks!
> > >
> > > Best,
> > >
> > > Wen
> > > *
> > > *   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/
> *
> *   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/