November 1 Lecture:
Kahn, H Intro to epidemiologic methods 1985 readings for today’s class
- Going to make challenging part a bonus
Nov 1 :
Formula and source: OR calculator
(Refer to PowerPoint slides posted oct 30 “complications…”)
A + B C
A= risk factor or exposure factor
Confounding or effect modification
Confounding: Is something you want to avoid
Effect modification: Is a little more informative, we can conceive of this as saying
there is a relationship between a and b
Categories such as age groups are what we refer to as strata or levels.
Relationship of a and b changes depending on which level or strata you are
i.e smoking when you are 50 as opposed to being 15.
Effect modification is : Are the 3 strata homogeneous? Is the effect the same?
If we compute an OR for one strata is it 1? How about for the next strata?
Strata could be age (physiological), sociological, you can have different strata
with different ethnicities (can be testing intrinsically genetics or diets)
Effect modification is a really good tool to see if the relationship between a and
be stay constant.
Assess interactions mathematically
primary risk factor for having a heart attack is hypertension
Effect modification then is AGE
Ordinary chi square/ complicate chi square
Null hyp: no association between blood pressure and heart attack
Alternative: there is Slide 6:
Contingency chi square does not depend on sample size
MH chi square:
Only 2 by 2
Screw this table it is wrong
1 degree of freedom
3.84= critical value
We can conclude that age and heart attacks are associated. They are
OR value: 3.44
4>1.67 therefore there is signifcantly more people having heart attacks.
Adjusted residuals: statistical way to figure out whether or not that cell is
If adj residual > 1.67 ; there are significantly more… than expected than chance.
<1,67 tells us there are significantly less than expected.
This is the proper way of stating odds ratio.
This is the template for OR value statement (assignment 3).
Proper values will be given next week.
Slide 12: OR>1, then there is an association
Chi square is