Charles River Case

It is important to determine whether there is enough information using the variables constructed in this current model, or if more need to be added from the data set. For this logistic regression, Chi-square is not significant. This maintains that introducing additional variables would gain very little, if any at all.

The odds ratio shows us the probability of these variables having an effect on purchasing Florence vs. not having an effect. By looking at this, we can tell how much of a factor each variable is in deciding whether or not someone would by Florence or not. For instance, the data suggests it is significantly more likely that a female would by this book compared to males. The two biggest deciding variables for purchase are:

1. Being a book club member

2. Frequent purchaser of Youth Books

Misclassification Rate- 301/2600 = 11.58% Accuracy- 2299/2600= 88.4%

Accuracy for buyers- 32/307= 10.4%

% customers to whom offers made-58/2600=2.2%

* 10,000 customers, mailed out offers to 220

% books bought- 32/2600= 1.2%

* 10,000 customers, 120 will have bought the book, and 220-120= 100 retuned to ADMC

ADMC Variable only

I thought it would be interesting to run a model using only ADMC provided variables. Maybe this could shed a different light that was missing from original model. Right off the bat we can see the significant value of the chi square ruling this model out. More variables are needed to make a better predictor on who is going to purchase Florence or not. Looking further, it seems these variables are unimportant in comparison to the data provided by CRB.