ANTC67 – Lec 8
•eg smoking (exposure factor) is a medical condition
•What would expose people to a certain entity that would decrease or increase rate
of smoke?
•Lung cancer (disease condition)
•Eg autism (condition)
•What exposure factor is considered to give rise to the condition
•Vaccine
•Eg diabetes (condition)
•What exposure factor contributes to elevation of diabetes
•Obesity
•Eg breast cancer
•Eg estrogen pills
•Breast feeding could lower the rate of breast cancer
•Have matrix
•Always +/- (disease present or not)
•Then exposure (yes and no)
•Then have high middle low income classes?
•Odds ratio: (a*b / b*c)
•If the value is around 1 then no relationship, no association
•But before can interpret what Odds ratio is, we have to look at Chi Square
•Because chi square says ‘highly sig’ means that this relationship is stsistically sig
•Eg breast cancer (y/n), child (y/n), breastfeed (y/n), how many children (y/n) (across)
•Names of women going down
•Hypothesis is basically the question
•Null hypothesis (Ho) (no relationship)
•Is there a relationship, which is H1
•So we’ll be using chi square to test our odds ratio
•Region of rejection: how much error are you willing to let into your decision
•How confident are you that you made the right decision
•Alpha is typically 0.5, so 0.5% error, and we allow this because we are always dealing with
samples
•Degrees of freedom= (rows-1)*(columns-1)
•If in the assignment there is a percentage, convert that percentage into a whole number
•No cell in matrix should be zero
•But can put in decimal points?
•MH chi square: MH stands for Mantel and Haenezel
•Formulas are on the exam
•Common rule of thumb is that in a b c d should not drop below 5, so staticians let it even
go down to 1
•