## 2.5 Other Considerations

The approach presented here is not the only approach that could have been taken with these data. For example, if logistic regression is the model being evaluated, the `glmnet`

model (Hastie, Tibshirani, and Wainwright 2015) is a model that incorporates feature selection into the logistic regression fitting process^{14}. Also, you may be wondering why we chose to preprocess only the imaging predictors, why we did not explore interactions among risk predictors or between risk predictors and imaging predictors, why we constructed interaction terms on the original predictors and not on the preprocessed predictors, or why we did not employ a different modeling technique or feature selection routine. And you would be right to ask these questions. In fact, there may be a different preprocessing approach, a different combination of predictors, or a different modeling technique that would lead to a better predictivity. Our primary point in this short tour is to illustrate that spending a little more time (and sometimes a lot more time) investigating predictors and relationships among predictors can help to improve model predictivity. This is especially true when marginal gains in predictive performance can have significant benefits.

### References

Hastie, T, R Tibshirani, and M Wainwright. 2015. *Statistical Learning with Sparsity*. CRC press.