First, it does not mean "control for" in the English sense. This suggests that you do something to fix the variable. The procedure is not doing that. The procedure is saying that if you are predicting income, and have two people with the same years of education, that the model predicts the person working more years has a higher income (assuming the coefficients are all positive). Multiple regression is usually covered in the second half of introductory statistics books (and there are lots of specialist regression books too). I recommend looking at some of these as they will give more thorough answers than what you will likely get here.

First, it does not mean "control for" in the English sense. This suggests that you do something to fix the variable. The procedure is not doing that. The procedure is saying that if you are predicting income, and have two people with the same years of education, that the model predicts the person working more years has a higher income (assuming the coefficients are all positive). Multiple regression is usually covered in the second half of introductory statistics books (and there are lots of specialist regression books too). I recommend looking at some of these as they will give more thorough answers than what you will likely get here.

Dear  Witold:

From a mathematical or statistical point of view, to control the impact of one or more predictors we do one of the following:

1. Introduce some knowledge about this predictors in your model in the form of a constraint such as a fixed constraint.

2. Assign different weights in the weight matrix for each predictor.

As such you can control or limit the impact of your predictors in your mathematical model. We assume that we are using some sort of least squares optimization.

Regards,

Gamal

Hi,

when there is an extraneous variable we want to exclude its effect, we do control on this variable. unles you know this variable and dealth with it as covairate, you dont need to write the statement in your question.

usually we do hierrarchical regression to control for the covariate.

Regards

Witold, multiple regression takes into account the covariance of the independent variables when estimating the slope coefficients, attempting to estimate the unique effect of each variable, independent of the other variables in the model. In your example years in education and years of working experience may be correlated. Then, the regression coefficients (slopes) will change compared to regression analysis in which only one of these variables is included.

Here you will find more information (under the heading Unique Prediction and Partial Correlation):

http://www.statsoft.com/Textbook/Multiple-Regression#cunique

HTH, Karin

In multiple regression analysis, suppose there are two independent variables, will get two partial coefficients and in interpretation of these coefficients, one will give the change in dependent variable for unit change in that independent variable and other variable is kept constant.  

Thanks t everybody for explanations. They were very helpful. 

It is wrong to say  "we control for another variable"  .  You should say we control the effect of  other variable(s) which means we remove the effect of other variables from the relation between the two or more variables, this implies that we keep the effect of other variables brought explicitly in the model constant. To elaborate and expand check residual analysis in my book:

Applied Econometrics using Eviews, SPSS & Excel with Applications in Arab Countries (2012). Available at AMAZON. First edition 10/2011; AuthorHouse Publishing (October 19, 2011)., ISBN: ISBN-10: 1456775790 ISBN-13: 978-1456775797

http://www.amazon.com/s/ref=nb_sb_noss?url=search-alias%3Dstripbooks&field-keywords=Charbaji