Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

Re: st: control a variable in stata

From   Nick Cox <>
To   "" <>
Subject   Re: st: control a variable in stata
Date   Sat, 21 Apr 2012 13:54:23 +0100

The research question is explaining salaries. That being so you would be better off with -poisson- or -glm, link(log).

OLS is an estimation method, not a model.


On 21 Apr 2012, at 13:33, "Kong, Chun" <> wrote:

Dear Nora,

Thank you very much for your advice!! I really appreciate for your time and help :) I am going to add a race and age variable and see how they affect on player's salary. I have look through the paper you have suggested and other studies with the related topic and they gave me many great ideas!! Regarding the choice of model, do you mean that OLS is the appropriate and the only model I should if I only have data in 1 season??

Thank you very much for your help again!

Yours sincerely
From: [owner-] on behalf of Nora Reich []
Sent: 20 April 2012 17:15
Subject: Re: st: control a variable in stata

Dear Andy,

first some ideas about your independent variables:

1. You distinguish between players born in the US and players born
outside the US. I would suggest to also control for skin colour (black
or white), either only for those born in the US or for all (depending
on the results of these estimations), because skin colour seems to
affect the salary as well, see, for example, this paper: (which I found
by simply googling).

2. Another important factor might be the number of years the player
has played in the NBA.

3. Maybe age also plays a role?

Generally, my advice would be to look at papers with a similar
research question and derive your list of independent variables from
the literature review (and, of course, from own ideas). Have you done
a literature review? This is usually a good thing to do before
estimating regressions.

Random effects and fixed effects models are for panel data. If you
have only 1 NBA season, these models are not appropriate. However, if
you have a variable "year" which tells you whether the data is from
2010 or 2011, it would be valuable to include a dummy for one of the
years in your regression.

For the tests for the assumptions of the OLS model, just google
something like "regress postestimation stata".

Best regards

Am 20. April 2012 16:11 schrieb Kong, Chun <>:
Dear statalist,

I am working on a paper in finding the determinants of NBA players' salary. Data are collected from the 2010-2011 NBA season. My dependent variable is ln(salary). I have got several dummy variables

1) ethnicity (0 if player is born in US, 1 for international player)
3)Season Played in the NBA
3)Efficiency Index
5)Approximate Value Index
6)Versatility Index
7)Points per Field Goal
8)Turnover to assist Ratio

My results turn out that the salary of international player is higher relative to the players who born in US. However, to make the comparison fair, I want to test the effect of ethnicity on player's salary while controlling the performance of both international players and US players.

At the moment, I am now only working on a simple OLS model. May I ask for your advice that what can I try or do to make my results better? ( I have read something like the random effect and fixed effect model, but I am really not sure what I can do). Also, do I need to do some tests to check the problem such as endogeneity in my model

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index