Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

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

Re: st: peer variable coefficient estimate nonsense

From   Nick Cox <>
Subject   Re: st: peer variable coefficient estimate nonsense
Date   Thu, 1 Mar 2012 08:32:01 +0000

What is your story then explaining the _negative_ correlation? Have
you plotted the data in various ways to explore what is going on?

See also David Hoaglin's wise words at


On Thu, Mar 1, 2012 at 8:01 AM, Jian Zhang <> wrote:
> Nick, thanks for your comment.  the correlation between age and peer
> average age is quite high, about -0.7
> On Thu, Mar 1, 2012 at 3:56 PM, Nick Cox <> wrote:
>> You need to think very carefully how to parameterise this. You don't
>> give any comment on how far these age variables are correlated. You
>> might be better off with a different representation, e.g. predictors
>> of
>> age, (peer average - age)
>> Nick
>> On Thu, Mar 1, 2012 at 7:45 AM, Jian Zhang <> wrote:
>>> I was running a OLS regression of student scores on her/her own age
>>> and his/her peers' average age. The results show that the coefficient
>>> estimate of the peer average age is much much larger than own age
>>> estimate.  This does not not seem to be plausible as one would believe
>>> that own age would matter much more than peer average age to own
>>> scores.  any thoughts/comments?  looks like whenever adding peer
>>> characteristics, the regression results does not make sense!

*   For searches and help try:

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