+ Reply to Thread
Results 1 to 3 of 3

Thread: "Adjusting" logistic regression model

  1. #1
    Points: 6,655, Level: 53
    Level completed: 53%, Points required for next Level: 95

    Posts
    44
    Thanks
    0
    Thanked 0 Times in 0 Posts

    "Adjusting" logistic regression model




    I'm trying to predict response to a direct mail campaign, but there will be multiple campaigns, each having a different "topic" (eg, home remodeling, car repair, etc). We think topic has bearing on the response so we want to "adjust" the model accordingly.

    Options I see:

    1) Create a distinct model for each topic. We want to explore other options first.

    2) Interactions. For my training data, I would use multiple campaigns of different topics. I detect an interaction between topic and age (eg, for home remodeling older customers have higher response but for car repair younger customers have higher response). So when I am ready to score, this model would take into account the topic of the upcoming campaign and output a score file that has "adjusted" for that topic.

    3) Build a "core" model to find the predictors. Then re-train it using the same predictors but a different campaign (having the topic of interest), thus producing different parameter estimates.

    Are there other approaches appropriate for this scenario? Thanks!

  2. #2
    Fortran must die
    Points: 58,790, Level: 100
    Level completed: 0%, Points required for next Level: 0
    noetsi's Avatar
    Posts
    6,532
    Thanks
    692
    Thanked 915 Times in 874 Posts

    Re: "Adjusting" logistic regression model

    They will probably have an impact on response rate which will further complicate your analysis when comparing results. The lower the response rate the less sure you can be of your results.

    How will you adjust for topic? If its an IV then the regression (if that is what you are doing) will show you if its a signficant variable and having it in the model will control for its effect. I am not sure what interaction term you are suggesting here.

    Your third solution seems best me off hand. Again if topic is a predictor (rather than something you predict) placing it in the regression should correct this problem. What it won't do at all is bring you the same response rates - ones with more interest will likely generate a higher response rate.
    "Very few theories have been abandoned because they were found to be invalid on the basis of empirical evidence...." Spanos, 1995

  3. #3
    Points: 6,655, Level: 53
    Level completed: 53%, Points required for next Level: 95

    Posts
    44
    Thanks
    0
    Thanked 0 Times in 0 Posts

    Re: "Adjusting" logistic regression model


    Quote Originally Posted by noetsi View Post
    How will you adjust for topic? If its an IV then the regression (if that is what you are doing) will show you if its a signficant variable and having it in the model will control for its effect.
    Yes, "topic" would be an IV. But I suppose I'm making an assumption that it would be a significant predictor in the first place. Let's just say that's true for this purpose.

    Quote Originally Posted by noetsi View Post
    Your third solution seems best me off hand. Again if topic is a predictor (rather than something you predict) placing it in the regression should correct this problem.
    Thank you for that vote.

    More thoughts welcome!

+ Reply to Thread

           




Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts






Advertise on Talk Stats