Controlled variables in logistic regression

Hi everyone,

I have major problem in designing logistic regression analysis in my master thesis. I’ve been working with large panel data related to companies in stock market. My aim is to understand organizational impacts affecting companies’ behavior on disclosing sustainability information. As dependent variable I used a dummy, whether the company published a sustainability report. Since my only aim is to understand organizational dynamics, I want to control anything related to finance of companies. However, I have no idea how I can I do this in logistic regression analysis with Python. Without controlling financial variables, the results became biased. Can anyone teach me how to do this?

Hello @korlu!
Just to make sure I understand what you’re trying to do, let me try to paraphrase here: You’d like to use logistic regression to determine which variables contribute a company disclosing sustainability information or not. You’d like to control for variability relating to companies’ finances. Is this correct?

If so, can you tell me a little bit more about the dataset that you’re working with? In addition can you tell me briefly how your results are biased when you use financial information?

Hi @microrunner33748,

Thanks for the reply. Great questions! I collect my own data and build the data set. In the literature, researcher suggests that financial data should be controlled in the analysis due to the bias (again this is also stated in the literature). Sorry if the answers are enough for you but this is the only way I can reply without all the theoretical details in the management literature.

Thanks everyone. I found the answer in other corners of the internet. For interested others, control variables should be added to the regression as independent variables. The difference between these two is not a statistical problem (so cannot solve with the software packages), but rather is a theoretical problem. Simply the estimates of control variables should be interpreted differently.