Can anyone answer the question or direct me to the proper resources - ESPECIALLY for sampling error effect on the coefficients? How would the coefficients be affected if the independent variables came from one data set while the dependent variable came from a different data set?

  • 1
    $\begingroup$ How can it be possible for the IVs and DVs to come from "separate" datasets when they are necessarily matched, one-to-one, in order to perform regression? Please explain more about the context of your question. $\endgroup$
    – whuber
    Sep 25, 2013 at 14:48
  • 1
    $\begingroup$ @whuber I wonder if user30682 is talking about fusing different secondary datasets. For instance, taking some county-wide health data from CDC and merge it with Census' 2010 county data for analysis, etc Since both datasets have their own sampling errors, the challenge is then how the analyst should evaluate the merged data's sampling error. $\endgroup$ Sep 25, 2013 at 18:29

1 Answer 1


Since what you are most interested in seems to be sampling error, I will try to answer this part part of your question. (You might want to consider editing your question and making three questions out of it.)

Sampling Error can lead to estimating coefficients, that fit your sample perfectly, but don't hold in a different sample. This phenomenon is also kown as overfitting. To check if overfitting did occur, you can estimate the parameters from one sample and then try to validate them with another sample. This is called cross validation (consider the name of this site for a second!).

In principle there is no problem with taking dependent variables from one dataset and independent variables from another, as long as both have the same number of data points. If it's a good idea or not depends on what exactly it is that you are trying to do.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.