I am doing enterprises credit default prediction, so I have independent variables including financial data and the NEWS. I mean the bad news of the enterprise when I mentioned news, if there was a bad news like the founder of the enterprise committed insurance fraud, I would expect that enterprise has a little higher probability having credit default, just my own idea, no offense.

The problem is, the financial data are relatively regular, they publish financial data quarterly, while the news is definitely not! News is kind of "RARE", you could get the observation of the bad news only when there is bad news. So, the variable NEWS is much sparse when compared with financial data. I am planning to establish a panel data model, do you think this SPARSE NEWS variable would lead to biased estimate? if the coefficient of the news variable is biased, would the coefficients of the financial data be biased?

Thank you so much, I kind of stuck in this point and can not proceed on my paper..

  • 1
    $\begingroup$ Distribution for predictors in logistic regression is usually not important, because you're building a model conditional on their values. $\endgroup$
    – SmallChess
    Commented May 9, 2017 at 3:24
  • 1
    $\begingroup$ Maybe stats.stackexchange.com/questions/67078/… ? $\endgroup$
    – SmallChess
    Commented May 9, 2017 at 3:25


Your Answer

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

Browse other questions tagged or ask your own question.