Applied Longitudinal Data Analysis for Epidemiology: A Practical Guide

by Jos W. R. Twisk

• ISBN13: 9780521525800
• ISBN10: 0521525802
• Edition: 1st
• Format: Paperback
• Copyright: 2003-04-28
• Publisher: Cambridge University Press

Summary

The most important techniques available for longitudinal data analysis are discussed in this book. The discussion includes simple techniques such as the paired t-test and summary statistics, but also more sophisticated techniques such as generalized estimating equations and random coefficient analysis. A distinction is made between longitudinal analysis with continuous, dichotomous, and categorical outcome variables. This practical guide is especially suitable for non-statisticians and all those undertaking medical research or epidemiological studies.

Author Biography

Jos W. R. Twisk is senior researcher and lecturer in the Department of Clinical Epidemiology and Biostatistics and the Institute for Research in Extramural Medicine, Vrije Universiteit, Medical Centre, Amsterdam.

Table of Contents

Preface	xv
Acknowledgements	xvi
Introduction	1	(6)
Introduction	1	(1)
General approach	2	(1)
Prior knowledge	2	(1)
Example	3	(2)
Software	5	(1)
Data structure	5	(1)
Statistical notation	5	(2)
Study design	7	(11)
Introduction	7	(2)
Observational longitudinal studies	9	(6)
Period and cohort effects	9	(4)
Other confounding effects	13	(1)
Example	14	(1)
Experimental (longitudinal) studies	15	(3)
Continuous outcome variables	18	(37)
Two measurements	18	(3)
Example	20	(1)
Non-parametric equivalent of the paired t-test	21	(2)
Example	22	(1)
More than two measurements	23	(14)
The `univariate' approach: a numerical example	26	(3)
The shape of the relationship between an outcome variable and time	29	(1)
A numerical example	30	(2)
Example	32	(5)
The `univariate' or the `multivariate' approach?	37	(1)
Comparing groups	38	(7)
The `univariate' approach: a numerical example	39	(2)
Example	41	(4)
Comments	45	(1)
Post-hoc procedures	46	(2)
Example	47	(1)
Different contrasts	48	(4)
Example	49	(3)
Non-parametric equivalent of MANOVA for repeated measurements	52	(3)
Example	53	(2)
Continuous outcome variables -- relationships with other variables	55	(47)
Introduction	55	(1)
`Traditional' methods	55	(2)
Example	57	(3)
Longitudinal methods	60	(2)
Generalized estimating equations	62	(15)
Introduction	62	(1)
Working correlation structures	62	(4)
Interpretation of the regression coefficients derived from GEE analysis	66	(2)
Example	68	(1)
Introduction	68	(1)
Results of a GEE analysis	69	(3)
Different correlation structures	72	(3)
Unequally spaced time intervals	75	(2)
Random coefficient analysis	77	(14)
Introduction	77	(1)
Random coefficient analysis in longitudinal studies	77	(3)
Example	80	(1)
Results of a random coefficient analysis	80	(8)
Unequally spaced time intervals	88	(1)
Comments	88	(3)
Comparison between GEE analysis and random coefficient analysis	91	(4)
Extensions of random coefficient analysis	92	(1)
Equal variances over time	92	(1)
A numerical example	93	(1)
The correction for covariance	93	(2)
Comments	95	(1)
The modelling of time	95	(7)
Example	98	(4)
Other possibilities for modelling longitudinal data	102	(18)
Introduction	102	(1)
Alternative models	102	(12)
Time-lag model	102	(3)
Modelling of changes	105	(2)
Autoregressive model	107	(1)
Overview	108	(1)
Example	108	(1)
Introduction	108	(1)
Data structure for alternative models	109	(1)
GEE analysis	109	(3)
Random coefficient analysis	112	(2)
Comments	114	(4)
Another example	118	(2)
Dichotomous outcome variables	120	(25)
Simple methods	120	(8)
Two measurements	120	(2)
More than two measurements	122	(1)
Comparing groups	122	(1)
Example	123	(1)
Introduction	123	(1)
Development over time	123	(3)
Comparing groups	126	(2)
Relationships with other variables	128	(17)
`Traditional' methods	128	(1)
Example	128	(1)
Sophisticated methods	129	(2)
Example	131	(1)
Generalized estimating equations	131	(6)
Random coefficient analysis	137	(3)
Comparison between GEE analysis and random coefficient analysis	140	(3)
Alternative models	143	(1)
Comments	144	(1)
Categorical and `count' outcome variables	145	(22)
Categorical outcome variables	145	(11)
Two measurements	145	(1)
More than two measurements	146	(1)
Comparing groups	147	(1)
Example	147	(4)
Relationships with other variables	151	(1)
`Traditional' methods	151	(1)
Example	151	(1)
Sophisticated methods	152	(1)
Example	153	(3)
`Count' outcome variables	156	(11)
Example	157	(1)
Introduction	157	(1)
GEE analysis	158	(5)
Random coefficient analysis	163	(2)
Comparison between GEE analysis and random coefficient analysis	165	(2)
Longitudinal studies with two measurements: the definition and analysis of change	167	(12)
Introduction	167	(1)
Continuous outcome variables	167	(8)
A numerical example	171	(2)
Example	173	(2)
Dichotomous and categorical outcome variables	175	(2)
Example	175	(2)
Comments	177	(1)
Sophisticated analyses	178	(1)
Conclusions	178	(1)
Analysis of experimental studies	179	(23)
Introduction	179	(2)
Example with a continuous outcome variable	181	(14)
Introduction	181	(1)
Simple analysis	182	(2)
Summary statistics	184	(1)
MANOVA for repeated measurements	185	(1)
MANOVA for repeated measurements corrected for the baseline value	186	(2)
Sophisticated analysis	188	(7)
Example with a dichotomous outcome variable	195	(5)
Introduction	195	(1)
Simple analysis	195	(1)
Sophisticated analysis	196	(4)
Comments	200	(2)
Missing data in longitudinal studies	202	(23)
Introduction	202	(2)
Ignorable or informative missing data?	204	(1)
Example	205	(2)
Generating datasets with missing data	205	(1)
Analysis of determinants for missing data	206	(1)
Analysis performed on datasets with missing data	207	(5)
Example	208	(4)
Comments	212	(1)
Imputation methods	213	(10)
Continuous outcome variables	213	(1)
Cross-sectional imputation methods	213	(1)
Longitudinal imputation methods	213	(1)
Multiple imputation method	214	(2)
Dichotomous and categorical outcome variables	216	(1)
Example	216	(1)
Continuous outcome variables	216	(3)
Dichotomous outcome variables	219	(2)
Comments	221	(2)
Alternative approaches	223	(1)
Conclusions	223	(2)
Tracking	225	(16)
Introduction	225	(1)
Continuous outcome variables	225	(5)
Dichotomous and categorical outcome variables	230	(4)
Example	234	(4)
Two measurements	235	(2)
More than two measurements	237	(1)
Comments	238	(2)
Interpretation of tracking coefficients	238	(1)
Risk factors for chronic diseases	239	(1)
Grouping of continuous outcome variables	239	(1)
Conclusions	240	(1)
Software for longitudinal data analysis	241	(39)
Introduction	241	(1)
GEE analysis with continuous outcome variables	241	(6)
STATA	241	(2)
SAS	243	(1)
S-PLUS	244	(2)
Overview	246	(1)
GEE analysis with dichotomous outcome variables	247	(3)
STATA	247	(1)
SAS	248	(1)
S-PLUS	249	(1)
Overview	250	(1)
Random coefficient analysis with continuous outcome variables	250	(13)
STATA	250	(1)
SAS	251	(4)
S-PLUS	255	(2)
SPSS	257	(2)
MLwiN	259	(3)
Overview	262	(1)
Random coefficient analysis with dichotomous outcome variables	263	(8)
Introduction	263	(1)
STATA	264	(1)
SAS	265	(4)
MLwiN	269	(1)
Overview	270	(1)
Categorical and `count' outcome variables	271	(1)
Alternative approach using covariance structures	272	(8)
Example	274	(6)
Sample size calculations	280	(6)
Introduction	280	(3)
Example	283	(3)
References	286	(9)
Index	295