The University of Hong Kong
Department of Community Medicine
School of Public Health
Master of Public Health
Advanced Statistical Methods II Analysis of complex data
Practical 3
(10:00am 1:00pm July 2, 2011)
These files (/home/asm2/s7/pra3wide.dta, /home/
Practical 2 solution
A
1. We have a given you data that includes many measurements for each child.
Measurements for a particular child are likely to be correlated, so this is clustered
data. Analysis of clustered data needs to take into account the correl
The University of Hong Kong
Department of Community Medicine
School of Public Health
Master of Public Health
Advanced Statistical Methods II Analysis of complex data
Practical 2
(10:00am 1:00pm June 18, 2011)
This is a longitudinal survey of 880 children,
Practical 1 solution
. use /home/asm2/s3/pra1.dta,clear
. describe
obs:
2,895
vars:
12
10 Jun 2011 12:07
size:
95,535 (99.9% of memory free)
-storage display
value
variable name
type
format
label
variable label
-age
byte
%8.0g
age
sex
byte
%8.0g
sex
sex
e
The University of Hong Kong
Department of Community Medicine
School of Public Health
Master of Public Health
Advanced Statistical Methods II Analysis of complex data
Practical 1
(10:00am 1:00pm June 11, 2011)
This is a cross sectional survey of people fro
The University of Hong Kong
Department of Community Medicine
School of Public Health
Master of Public Health
Advanced Statistical Methods II Analysis of complex data
Assignment 5
(submitted to WebCT on or before 12:00pm, July 4, 2011)
June 29, 2011
Please
The University of Hong Kong
Department of Community Medicine
School of Public Health
Master of Public Health
Advanced Statistical Methods II Analysis of complex data
Generalized estimating equations
Assignment 4
(submitted to WebCT on or before 12:00pm, J
Assignment 3 key points
1.
How many levels are used in the model?
There are 2 levels: individual level and country level
2.
Please identify all the variables at each level.
Individual level: Age, Family Affluence Scale (FAS), sex, smoking behavior
Country
The University of Hong Kong
Department of Community Medicine
School of Public Health
Master of Public Health
Advanced Statistical Methods II Analysis of complex data
Multi-level model
Assignment 3
(submitted to WebCT on or before 12:00pm, June 20, 2011)
J
The University of Hong Kong
Department of Community Medicine
School of Public Health
Master of Public Health
Advanced Statistical Methods II Analysis of complex data
Mediation
Assignment 2
(submitted to WebCT on or before 12:00pm, June 15, 2011)
June 9, 2
Papers
Contributors: JD and CP conceived the study. All, in particular
MD, developed the methods. MD was responsible for the analysis and producing the maps. JD and CP wrote the article. MD
commented critically on the paper. CP revised the paper in the
li
Assignment 1 Part II
Some common mistakes
Did not provide STATA command/programme
Did not interpret STATA output
Misunderstood _merge=1 or =2
Misunderstood row/column percentage in Q4
Did not use logistic regression in Q7
Interpreted odds ratio OR as rela
The University of Hong Kong
Department of Community Medicine
School of Public Health
Master of Public Health
Advanced Statistical Methods II Analysis of complex data
Model building and selection, interaction
Assignment 1
(Part I bring it to next session,
Advanced statistical methods
II
Review
Practical 3
Course review
More examples of how to tackle complex
longitudinal data
Answers for this practical
Types of data
Individual
Multi-level or clustered or structured
Longitudinal
Modelselection
Typicalp
Regression of BP on all weight
z-scores
.xi:regressbpsysi.sexwz0wz3wz12wz96wz132
i.sex_Isex_12(naturallycoded;_Isex_1omitted)
SourceSSdfMSNumberofobs=7774
F(6,7767)=372.18
Model195842.863632640.4772Prob>F=0.0000
Residual681164.834776787.6998628Rsquared=0.
Advanced Statistical Methods
II
Multi-level models
Course structure
Ways of thinking about and tools for handling
1. Inter-relationships within individual data usually relating
to one or two time points, e.g., cross-sectional survey,
case-control study, c
Advanced statistical methods
II
GEE
Course content
Inter-relations considering individual level
attributes from one or two time points
Inter-relations considering individual and
group level attributes
Inter-relations taking into account
repeated measur
CTCE 6040 Advanced statistical methods II Analysis of complex data
Coordinator: Dr M Schooling
Module Description:
Introduction to Stata , analysis of complex and longitudinal data.
Prerequisite: Statistical methods
Summer Semester (Concentrated: 8 June -
Advanced statistical methods
II
Complex longitudinal
data
Learning objectives
Recognize complex longitudinal data
Appreciate the need for and complexity of
handling such longitudinal data
Be aware of the strengths and
weaknesses of various methods of
hand
Advanced statistical methods
II
Mediation & PathAnalysis
Learning objectives
Implement an analytic strategy using
mediation
Implement an analytic strategy using path
analysis
Appreciate the role of structural equation
modelling
stata usage
Menu or comm
Advanced statistical methods
II
Dr CM Schooling
Cindy Lin
Coursestructure
Lectureswithexamplesandquestions
Assignments
Practicalsessions
Coursecontext
Weareusuallytryingtomakecausalinferences
forpublichealthpolicyandlifestyle
recommendations,e.g.,
Opti
Your turn
Is mothers birthplace associated with
childs overweight status (ov)?
use /home/asm2/s5/wide130,clear
Logistic regression at 7 years
xi:logistic ov7 i.hkm i.sex
.xi:logisticov7i.hkmi.sex
i.hkm_Ihkm_12(naturallycoded;_Ihkm_1omitted)
i.sex_Isex_
The University of Hong Kong
School of Public Health
CTCE 6020 Advanced Statistical Methods I
TUTORIAL 3 Suggested solution
Question 1: Radioactivity in vegetables
a)
Construct the X-chart and S-chart for the radioiodine level in the first 52 weeks with
co
Objectives
Experimental design
After this session, students should be able to
CTCE6020 Session 3
Perform and interpret ANOVA test
Recognize F distribution and interpret F test
Eric Lau
Recognize different experimental designs
School of Public Health
Th
Objectives
Estimating effects: Alternative
approaches to reduce confounding
After this session, students should be able to
Reduce bias in estimating treatment effect in
CTCE6020 Session 4
nonexperimental study, using propensity score method
Estimate tre
The University of Hong Kong
School of Public Health
CTCE 6020 Advanced Statistical Methods I
TUTORIAL 4 Suggested solution
Analysis of patient satisfaction
a)
What is the sample size?
388
b)
How many subjects are included for the factor analysis?
[Analyze