General Handouts and Reference Material on Biostatistical Modeling

• Syllabus from a course on Biostatistical Modeling(updated 9Jan03; click here for a pdf file for printing), lecture notes, concepts to master and supplemental notes (6 pages; updated 26Jan99)
• Homework assignments
• Problems with categorizing continuous variables
• Some useful equations for linear regression (26Jan99, PDF, 3 pages)
• Handouts on ordinary least squares regression and regression diagnostics by Jim Patrie (20Mar01, PDF, 73 pages)

• S scripts demonstrating various modeling building blocks. Note: If Windows S-Plus is open when you click to open one of these files, the script will be placed in the S-Plus script editor window automatically. Before clicking F10 to run the scripts, make sure any open GraphSheet is on top. Window ... Tile Vertical works well.
  • Curve fitting for a single predictor: least squares, least sum of absolute deviations, least median squared deviation, loess (requires Hmisc library, updated 26Jan99)
  • Fitting linear, linear spline, quartic polynomial, unrestricted cubic spline, and restricted cubic spline functions for a single predictor (requires Hmisc and Design libraries, 11Dec02)
  • Random examples of restricted cubic spline functions with 3 - 7 knots (requires Hmisc library, 25Feb99)
  • Miscellaneous examples of splines and effects of changing knots
  • Demonstration of loess details by Greg Snow, BYU
• Example of using restricted cubic spline functions to model time-response profiles from serially collected patient data (hormone levels from blood draws every 10 minutes in each subject)
  • Cluster bootstrap used to get confidence intervals accounting for intra-subject correlation
  • Bootstrap bumping used to get simultaneous confidence bands
  • Estimates and bands are for the difference in time-response mean profiles for older vs. younger subjects
  • Analyses done using R Hmisc package function rm.boot
• Kevin Quinn's web site on logistic regression and log-linear modeling (has good examples of the S glm function and some Bayesian methods)
• Binary logistic regression example from basic biomedical research showing a fit with and without interaction (Wenger et al, Can J Physiol Pharmacol 62:224, 1984)
• Ordinary Least Squares Regression and Regression Diagnostics (Jim Patrie, 20Mar01, 73 pages)
• Notes of problems in using ratios as dependent variables, from Rich Goldstein (15Jul99)
• Use of the Cox Semiparametric Regression Model for Predicting Costs, Charges, and Length of Stay; FE Harrell and LR Smith;Short Course, Meeting on Methodologic Issues in Health Services and Outcomes Research, 3Dec95)
• REGRESSION MODELING STRATEGIES with Applications to Linear Models, Logistic Regression, and Survival Analysis, Book and Short Courses by FE Harrell.
• MSCI Biostatistics II Course?
• R-based site for rank-based linear models by Abebe, Crimin, McKean, Haas, Vidmar (see Drug Information Journal 35:947-971, 2001).
• Doug Bates' presentation Using Open Source Software to Teach Mathematical Statistics, with examples of distribution fitting using maximum likelihood, profile likelihood plotting, and other statistical calculations
• University of Otago's student tutorials on Monte Carlo methods, principal components analysis, the bootstrap, clustering,

-- FrankHarrell - 09 Mar 2004, 27 Jun 2004, 4 Jul 2004, 14 Aug 2005, 7 Apr, 16 Apr 2008
-- ColeBeck - 09 Mar 2004