No. If the study design had been either case-control or
cross-sectional, risk estimates could not be computed
because the constant term (a) in the model could not
be estimated. In other words, even if the computer
printed out values of1.18 or1.19 for the constant
terms, these numbers would not be legitimate
estimates ofa.
For case-control studies, only odds ratios, not risks or
risk ratios, can be computed directly from the fitted
model.
ORc(SOC¼1vs.SOC¼0 controlling for SMK and SBP)
¼e^b,whereb^¼ 0 : 50 is the estimated coefficient of SOC
in the fitted model
¼exp(0.50)
¼0.6065¼1/1.65.
The estimated odds ratio is less than 1, indicating that,
for this data set, the risk of CVD death for high social
class persons is less than the risk for low social class
persons. In particular, the risk for low social class
persons is estimated as 1.65 times as large as the risk
for high social class persons.
Choice (a) isnotappropriate for the effect of SOC
using model 1. Model 1 contains interaction terms,
whereas choice (a) is appropriate only if all the
variables in the model are main effect terms. Choices
(b) and (c) are two equivalent ways of stating the
general formula for calculating the odds ratio for any
kind of logistic model, regardless of the types of
variables in the model.
