Defending the final model

- Now that we have decided we can't improve…our model any further, I'll talk to you about ways…to defend your final model.…So, the model did not fit very well.…But we think that's the best we can do.…Not everyone on your research team…will agree with that necessarily.…In my experience, I'm used to getting pushback…from people on my research team.…Often people who don't know statistics,…so it's particularly annoying.…However, I have come up with some good defenses…for the choices I make for my final model,…and I'm going to share them with you in this movie.…

First, sometimes they don't want any…nonsignificant covariates in the model.…Or barely significant covariates, like the ones that were…on the line just above p = 0.05.…If that happens, just take them out.…No reason to fight.…But tell them that you will have less to talk about…in the paper, which is true.…And remember, you can't take out the exposure,…but you can take out any other ones.…Next, sometimes they want the opposite.…Some extra nonsignificant stuff in the model…