Chapter 9
Estimating the Arithmetic Mean Difference
Sex group and blood pressure
The arithmetic mean difference was not my preferred measure of effect (chapter 3), but
for several related reasons I decided to give it a place of honor in the second part of
Estimating the Odds Ratio from
Matched Case-Control Studies
1
The whole truth about matched case-control studies
Matching controls to cases
does
NOT resemble matching unexposed to exposed
does
NOT eliminate confounding bias
superimposes
colliding bias on
Homework 5: Estimating the Modified Odds Ratio
(A cross-sectional sample)
In this homework, we are interested in several causes of an unnamed disease (disease).
The cross-sectional sample included 2,000 people. Relevant variables are shown below:
AGE (con
4/10/2012
Estimating the Rate Ratio
1
Setting: a cohort
N=15,712 people at baseline
cfw_
391 fell victim to ischemic stroke during an average follow
up of 10 years
Two causal variables
cfw_
cfw_
hypertension status (an exposure) - binary variable
age (a c
Homework 4: Estimating the Odds Ratio
(A cross-sectional study)
The data file for this homework was taken from a study of replacement hormones and
sleep apnea (Shahar et al. Am J Respir Crit Care Med 2003;167:11861192). Skim the
article (provided).
The in
Homework 8: Estimating the Hazard Ratio
(A cohort study)
The data were taken from of small cohort study of patients with severe skin cancer. A total of 65 patients
were followed for 1-77 months and 48 of them have died.
Variables:
Name
Description
Codes/R
Homework 4: Estimating the Odds Ratio
(A cross-sectional study)
The data file for this homework was taken from a study of replacement hormones and
sleep apnea (Shahar et al. Am J Respir Crit Care Med 2003;167:11861192). Skim the
article (provided).
The in
Homework 6: Estimating the Matched Odds Ratio
(An individually matched case-control study)
The data file for this homework was taken from a study of endometrial cancer (Mack et
al. New England Journal of Medicine 1976:294;1262-1267).
The investigators ide
Homework 5: Estimating the Modified Odds Ratio
(A cross-sectional sample)
In this homework, we are interested in several causes of an unnamed disease (disease).
The cross-sectional sample included 2,000 people. Relevant variables are shown below:
AGE (con
Homework 7: Estimating the rate ratio by tabular methods and by Poisson
regression
(Cohort, Group-level data)
To study the effect of a surgical procedure on mortality, a cohort of patients who underwent surgery was
compared with patients who did not.
Vari
Homework 3: Estimating the Modified Mean Difference
(A cross-sectional sample)
Gaining weight leads to an increase in waist circumference. In this assignment we are interested
in estimating that effect, and in possible modification of the effect by sex gr
Homework 1: instructions
The basic figure was taken from a short article (slightly modified). You might want
to skim the article, although it is not essential
Vlzke et al. Menopausal status and hepatic steatosis in a general female
population. Gut 2007;56
Estimating the Modified Mean Difference
1
Recall confounding and adjustment
Previous example: Regress SBP on age and sex
Exposure Confounder
SAS code:
proc glm;
model sbp = age sex /solution;
run;
Mean SBP = 79.8 + 0.735Age + 1.7Sex
The coefficient of age
Estimating the Odds Ratio
1
Introduction
When do we compute odds ratios (or probability ratios)?
cfw_
cfw_
How do we estimate odds ratios?
cfw_
cfw_
Effect is binary
No information on person-time at risk
Tabular analysis (contingency tables; cross-classif
Chapter 12
Estimating the Odds Ratio
Overweight and sleep apnea
Many otherwise healthy people do not breathe normally when they sleep, and most of
them are not even aware that something goes wrong every night. Instead of breathing
peacefully and orderly t
Estimating the Odds Ratio
1
Introduction
When do we compute odds ratios (or probability ratios)?
How do we estimate odds ratios?
Effect is binary
No information on person-time at risk
Tabular analysis (contingency tables; cross-classification)
Logistic re
Estimating the Hazard Rate Ratio
Introduction
Hazard (or Hazard Rate)
What is it?
The rate at the interval [t, t+t], when t0
A measure of the instantaneous chance of an event at time t
Not a probability!
Units: events/time (like rate)
Range: > 0 (like rat
Clinical Epidemiology
Dovepress
open access to scientific and medical research
M ethodolo g y
Open Access Full Text Article
Causal diagrams and the logic of matched
case-control studies
This article was published in the following Dove Press journal:
Clini
Chapter 10
Estimating the Modified Mean Difference
Sex group: an effect modifier?
At the end of the last chapter we tried to estimate the effect of age on systolic blood
pressure, assuming that the marginal association between the two variables contained
12/19/2012
Estimating the Modified Odds Ratio
1
Modified odds ratio
Similar to estimating the modified mean difference,
but
cfw_
The dependent variable is "log odds (Y=1)" rather than
"mean Y
cfw_
Regression coefficients: log odds ratios
cfw_
Exponentiate
Estimating the Modified Mean Difference
1
Recall confounding and adjustment
Previous example: Regress SBP on age and sex
Exposure Confounder
SAS code:
proc glm;
model sbp = age sex /solution;
run;
Mean SBP = 79.8 + 0.735Age + 1.7Sex
The coefficient of age
Estimating the Modified Odds Ratio
1
Modified odds ratio
Similar to estimating the modified mean difference,
but
The dependent variable is "log odds (Y=1)" rather than
"mean Y
Regression coefficients: log odds ratios
Exponentiated coefficients: odds ratio
Commentary
On effect modification and its applications
In a deterministic universe component causes join
hands to form a sufficient cause of an outcome. For
example, the mutated gene for phenylalanine
hydroxylase awaits the arrival of phenylalanine in the
Estimating the Rate Ratio
1
Setting: a cohort
N=15,712 people at baseline
391 fell victim to ischemic stroke during an average follow
up of 10 years
Two causal variables
hypertension status (an exposure) - binary variable
age (a confounder or an effect mo
12/19/2012
Estimating the Odds Ratio from
Matched Case-Control Studies
1
The whole truth about matched case-control studies
Matching controls to cases
does NOT resemble matching unexposed to exposed
does NOT eliminate confounding bias
superimposes collidi
5/26/2015
Estimating the Hazard Rate Ratio
Introduction
Hazard (or Hazard Rate)
What is it?
The rate at the interval [t, t+t], when t0
A measure of the instantaneous chance of an event at time t
Not a probability!
Units: events/time (like rate)
Range: > 0