18,221 reputation
12571
bio website biostat.mc.vanderbilt.edu/…
location Nashville, TN
age
visits member for 3 years, 3 months
seen 2 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.


2h
comment Non-significant factors after stepwise regression
What made you use stepwise regression? Do you know how to run simulation studies that demonstrate how poor these methods perform?
13h
revised Using empirical null distribution to adjust odds ratios
edited tags
13h
comment Using empirical null distribution to adjust odds ratios
Yes. You can find features using any algorithm, then estimate the bias by contrasting the apparent log odds ratio with the log odds ratio for that feature evaluated on the original sample. You could say that with sample with replacement the bootstrap sample is "super-overfitted" and the original sample us just "overfitted". The difference between super overfitted and overfitted mimics the difference between overfitted and non-overfitted, hence the justification for the bias correction.
16h
answered Using empirical null distribution to adjust odds ratios
17h
answered How to translate R to SQL for a Cox Proportional Hazards model?
18h
answered ROC vs. Accuracy
1d
comment How to apply predict functionality of R in SAS?
This is one of the needs that got me away from SAS and into the S language years ago - the ability to make predictions and get estimates using only the fit object. Before that I spent a lot of time converting SAS model estimates into Basic for offline evaluation. The SAS approach was to add test observations to the originall fitting dataset, setting $Y$ to missing for the test cases, and getting predictions for original + test cases together. Not very satisfying.
Jul
25
answered Is machine learning useful for comparing test group and control group
Jul
23
comment Power analysis for binomial data when sample sizes are different
You will do better in my opinion to state what you want to estimate, state the acceptable margin of error in estimating it, and then solve for $N$ that gives you that margin of error. The margin of error might be $\frac{1}{2}$ the width of the 0.95 confidence limit. The worst case margin of error is approximately obtained by setting $p=\frac{1}{2}$ and using $1.96 \sqrt(p(1-p)/n_{1} + p(1-p)/n_{2})$ where the two sample sizes are $n_{1}$ and $n_{2}$. You can set the ratio of the sample sizes to a constant $r$ and solve for one of the sample sizes, then the other.
Jul
23
answered Kruskal–Wallis Test not significant but Mann Whitney U significant
Jul
23
comment Power analysis for binomial data when sample sizes are different
Power is not that relevant after the data are collected. I would stick to a confidence interval for the difference in two probabilities.
Jul
22
comment Significance of varibles after stepwise regression
Along the same lines, one could say that if you are using stepwise regression why are you interested in inference at all?
Jul
21
answered How do you validate your machine learning models?
Jul
21
awarded  cross-validation
Jul
21
answered How to interpret coefficients of $x$ and $x^2$ in same regression
Jul
20
comment AIC, BIC and GCV: what is best for making decision in penalized regression methods?
See the code in the R rms package effective.df function and my book Regression Modeling Strategies. The main idea, from Robert Gray, is that you consider the covariance matrix without penalization vs. the covariance matrix with penalization. The sum of the diagonal of a kind of ratio of these two gives you the effective d.f.
Jul
20
answered AIC, BIC and GCV: what is best for making decision in penalized regression methods?
Jul
19
comment how to calculate effective degrees of freedom in ridge regression in R
Not for each variable but jointly for all the terms making up each variable.
Jul
19
answered how to calculate effective degrees of freedom in ridge regression in R
Jul
19
revised LASSO to identify important variables in ordered logistic regression?
edited tags