I am forecasting sales based on an e-mail data set. In the the dataset I have sales as well as the quantity of emails sent, the number of unique opens, unique clicks and unsubscribers. I have only two categorical values of type and category. I cleaned the whole dataset so it's by year and week.

I was requested to create a model (if possible linear) to predict sales by fiscal week. The old model I was shown used all the variables of quantity sent, unique opens, unique clicks and un-subscribers as coefficients for a multi-linear regression model. I don't agree with this since those variables are dependent on the quantity sent.

I was planning on creating a simple linear regression model of quantity sent predicting unique opens. Then creating a linear regression model to predict unique clicks based on unique opens. Then in my sales regression model, I was going to include unique clicks, the two categorical variables type and category and create a dummy variable for week (so if it's week 1 or 2 in the year etc).

The request wants to keep all of those variables [alhtough I am still doing a correlation analysis to make sure there is some kind of linear relationship] - so is this the correct approach? Thank You

  • $\begingroup$ Is your categorical variable binary? If so, you need to use a dummy or other valid contrast code. If it is not binary, is your categorical variable ordinal or nominal?. Based on the categorical data either OLS , GLM model has to be considered. $\endgroup$ – GD_N Mar 1 '17 at 5:28
  • $\begingroup$ I was going to create a dummy variable for the categorical variables, but I am nervous about a regression model based on explanatory variables (covariates) - since unique opens is directly related on quantity of emails sent. I don't believe I can use those two variables $\endgroup$ – jeangelj Mar 1 '17 at 17:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.