Notes

|  Cluster Analysis  |  Logistic Regression  |  Maximum Likelihood  |  Measurement Error Models  |  Missing Data  |  Modelling  |  Multicollinearity  |

|  Reliability  |  Repeated Measures  |  Simulation  |  Spatial Data Analysis  |  Bayesian Data Analysis  |

Cluster Analysis

• A page from the Clustan site which talks about missing values in cluster analysis.
• References: S-Plus 4.5 for Windows, Guide to Statistics, Chapter 18

Logistic Regression

• A page from a course in Quantitative Methods in Public Administration. Lots of good summary information on lots of issues in logistic regression.
• The PROPHET Statguide page on checking assumptions in logistic regression.
• Michael Friendly's page on Logistic Regression.
• Robert Wolfe's pdf file containing an explanation of logistic regression.
• Nan Lin's summary from the book "Logistic Regression Using The SAS System: Theory and Application" by Paul Allison
• A quote from Stata's help pages: "Conditional logistic analysis differs from regular logistic regression in that the data are stratified and the likelihoods are computed relative to each stratum. The form of the likelihood function is similar, but not identical, to that of multinomial logistic regression. Conditional logistic analysis is known in epidemiology circles as the matched case-control model and, in econometrics, as McFadden's choice model."

• Overdispersion

  • References:

• Sparse Data and Zero Cells

  • References: "Categorical Data Analysis" by Alan Agresti (Section 7.7)
• References: "Categorical Data Analysis" by Alan Agresti"Introduction to Categorical Data Analysis" by Alan Agresti"Applied Logistic Regression" by Hosmer and Lemeshow

Modelling

• From Agresti's "Introduction to Categorical Data Analysis" on the benefits of modelling relationships (when the model is correct): (i) You get to estimate effects while controlling for confounders (ii) You get to estimate the size of effects (as well as statistical significance (iii) You get predictions from the model

Measurement Error Models

• A statement of the problem.
• An article on Structural Equation Models and measurement error.
• An review article on measurement error and validation data.
• References: "Measurement error in nonlinear models", R.J. Carroll, D. Ruppert and L.A. Stefanski. (in Science Library, Call Number: 519.536 C31m)"Measurement Error Models", Fuller, W. A.
• Software: QVF beta-software for SAS and S-Plus (further development only being done for Stata?). Developed by Carroll among others.

Maximum Likelihood

• A page giving a simple example of maximum likelihood estimation.
• Doug Bates' JSM 2001 talk which includes examples of MLE and profile likelihood calculations.

Missing Data

• SOLAS homepage including description of single and multiple imputation.
• A multiple imputation FAQ
• S-Plus code for multiple imputation.
• Slides from a general talk on missing data.
• A page advertising the S-Plus 6.0 missing data library.
• An article (linked off the SOLAS site) which discusses the use of imputation in a large survey.
• Frank Harrell's "An Introduction to S-Plus and the Hmisc and Design Libraries" contains a section on imputation using the Hmisc library.
• Multiple Imputation On-line A site with all sorts of information and links on missing data analysis.
• References: "Statistical analysis with missing data", Roderick J.A. Little, Donald B. Rubin. (In the Science Library : 519.5 L77)"Analysis of Incomplete Multivariate Data", J. L. Schafer (Science Library: 519.535 s29)Alastair Scott (especially for missingness in survey data)
• Software: SOLAS will do it, but it costs money (!)Frank Harrell's Hmisc library will do it, but its not ported to R yet (although I must remember that I have S-Plus too!)There are the NORM, CAT and MIX libraries for S-Plus -- these are implementations from Schafer's bookSPSS will do it (including EM), but its harder to tell exactly what it is doing.

Multicollinearity

• A page discussing the problem (in a linear regression)
• NOTE: Multicollinearity is still a problem in non-Gaussian regression (e.g., logistic regression)

Reliability

• page including discussion and definition of Cronbach's Alpha.
• page on Item Response Theory.
• another page on Item Response Theory, including comparison with Classical Test Theory.
• page on Rasch Models.
• snippet on Item Response Theory.
• page comparing Item Response Theory and Classical Test Theory.
• section from StatSoft's Electronic Textbook on Reliability.x
• References: "The New Psychometrics: Science Psychology and Measurement", P. Kline"A Handbook of Test Construction", P. Kline"A Psychometrics Primer", P. Kline"Design and Development of Psychological Scales", C. Todd & C. Bradley"Health Measurement Scales", Streiner & Norman

Repeated Measures

• Don Hedeker's longitudinal analysis page with lots of pdf files explaining things.
• Don Hedeker's notes on Mixed Models for Longitudinal data.
• Andy Field's nice overview document on sphericity.
• A detailed example of a repeated measures analysis.
• Some excellent notes on mixed models in general and nlme in particular, by Doug Bates.
• A web page detailing how to perform and report the results of a repeated measures analysis (with one within-subjects factor) in SPSS.
• Hugh Foley's notes on The Single-Factor Within-Subjects Design.
• References: "Multivariate Analysis of Variance and Repeated Measures", Hand, D. J. and Taylor, C. C. (1987). Chapman and Hall. "Design and Analysis", Geoffrey Keppel "Mixed-Effects Models in S and S-Plus", Pinheiro and Bates

Simulation

• References: "Stochastic Simulation", B. D. Ripley

Spatial Data Analysis

• A paper on SAGE -- a software package for analysis of geographical data. Includes discussion of some of the issues with analysing spatial data.
• References: Venables and Ripley, Chapter 14Brian McArdle !

Bayesian Data Analysis

• Frank Harrell's Practical Bayesian Data Analysis from a Former Frequentist