Topics to discuss today during class 6/22/2016

[chenchen715]

1. Is Forward AIC a good approach to select significant variables? AIC of final selected model is not very different from the model that is selected one step ahead.
2. What will be a good way to test for colinearity? When looking at the final AIC selected model, it seems like WEEKEND, DRYOUNG and INTER_HWY are correlated.
3. Is p-value for coefficient estimation a reliable tool to understand if a predictor is significant? Heiberger mentioned that the coefficient significant is constructed under Type III ANOVA, we couldn't really find reference on it.
4. 3-way interaction, Patrick's riskScaleSimp does a great job in getting the odds for the 3-way interaction (time of day, gender and heavy_truck), we want to look at the odds together.
5. Once above three variables got recoded, the test for coeffieicnt significance changed dramatically, which should not really be the case.
6. Understand how to construct a Breslow-day test in complex survey.
7. How is Breslow-day test different from test for interation in logistic regression.

[nooreendabbish]

Great discussion today guys! I was wondering if either of you have experience plotting odds ratio CIs in R? I'm also simultaneously googling this...