19. Mortality: Multivariant Predictors

Now, wait a minute. The P value slipped to 0.06. Yes, when I excluded the 31% of patients in the study with diabetes in the multivariant model the risk that I am committing a type 2 statistical error increased to 6 out of 100 as compared to 3 out of 100 prior to this shift. In other words, there is a 6 % chance I am wrong. The other way to look at it is simple. Even in a smallish study (200 Patients), when I exclude 31% of the patients, atenolol still reduces the risk of death, WOW! That is a really, really, really, powerful effect.

Multi-variant Predictors

	If you exclude the patients with XYZ what do you see?
	Diabetes occurred in 31% of patients.
	If you exclude 31% of patients study has less power.
	6/100 vs 5/100 - Is that really different?

 

---