18,719 reputation
12773
bio website biostat.mc.vanderbilt.edu/…
location Nashville, TN
age
visits member for 3 years, 4 months
seen 11 hours ago

I am Professor of Biostatistics and Chairman of the Department of Biostatistics at Vanderbilt University School of Medicine, Nashville TN USA. I am Associate Editor of Statistics in Medicine, a member of the Faculty of 1000 Medicine, and a member of the policy advisory board for the Journal of Clinical Epidemiology. I am a Fellow of the American Statistical Association. I am author of Regression Modeling Strategies (Springer, 2001). My specialties are development and validation of predictive models, clinical trials, observational clinical research, cardiovascular research, technology evaluation, clinical epidemiology, medical diagnostic accuracy, biomarker research, pharmaceutical safety, Bayesian methods, quantifying predictive accuracy, missing data imputation, and statistical graphics and reporting. I am a long-time user of R. In August 2014 I was given the WJ Dixon Award for Excellence in Statistical Consulting by the American Statistical Association. Among many other things, Dr Dixon was the lead developer of the first general-purpose statistical software package, BMD.


11h
comment ROC for more than 2 outcome categories
How does the area under the curve not capture that?
22h
revised How to run main effects and interactions in a stepwise regression?
edited tags
22h
answered How to run main effects and interactions in a stepwise regression?
1d
comment ROC for more than 2 outcome categories
How does that give you any insight?
1d
comment ROC for more than 2 outcome categories
What about ROC curves makes them insightful to you? Are you really interested in concordance probabilities ($c$-index; ROC area; pure discrimination measure)? I find the ROC area to be helpful even though the curves are not helpful to me. And you can generalize the idea of concordance probability to multiple categories using Somers' $D_{xy}$ rank correlation coefficient.
1d
answered testing logistic regression coefficients using $t$ and residual deviance degrees of freedom
1d
comment Standard deviation of percentage changes in a time course
Absolutely. More information is at biostat.mc.vanderbilt.edu/ManuscriptChecklist under the heading Inappropriate choice of measure of change.
2d
answered Standard deviation of percentage changes in a time course
Aug
15
answered Can Mann-Whitney test be used for post-hoc comparisons after Kruskal-Wallis?
Aug
14
comment ordered logit-difference-in-difference
Get those from the coefficients (which represent difference from the reference cell in log odds scale) or by taking differences in coefficients to get other comparisons.
Aug
14
comment ordered logit-difference-in-difference
That sounds like a standard 2 degree of freedom test comparing 3 groups, which can be done by removing marital status from the model or by using a 2 d.f. contrast to get a Wald $\chi^2$ statistic. A double difference is not implied by the example question you provided.
Aug
14
answered Testing comorbidity adjusting for patient characteristics
Aug
14
revised Interpreting ordinal regression result and calculating individual percentage contributions of independent variables
edited tags
Aug
14
answered Interpreting ordinal regression result and calculating individual percentage contributions of independent variables
Aug
14
comment ordered logit-difference-in-difference
Please provide a specific simplified example of your desired result.
Aug
14
awarded  Good Answer
Aug
13
awarded  Nice Answer
Aug
13
comment Is it reasonable for a classifier to obtain a high AUC and a low MCC? Or the opposite?
No, you don't need classification unless you have a loss/utility function. Lookup lift curve for one method of using the results rationally - a cost/benefit analysis. Predicted risk also gives you the luxury of not classifying hard-to-classify observations.
Aug
12
comment Is it reasonable for a classifier to obtain a high AUC and a low MCC? Or the opposite?
It is good to start off using the correct terminology. If you can compute an AUC, the method you are using is not a classifier but rather is a risk (or other) predictor.
Aug
11
answered How to determine which variable or combination of the variables are affecting to the predictor variable?