25,875 reputation
141103
bio website biostat.mc.vanderbilt.edu/…
location Nashville, TN
age
visits member for 4 years
seen 27 mins 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.


Apr
20
awarded  Yearling
Apr
20
comment Multicollinearity in polynomial regression
From the standpoint of pure prediction, extreme algebraic co-linearities do not hurt.
Apr
17
awarded  Enlightened
Apr
16
comment Trust of coefficients of Logistic Regression
You don't really re-run the regression, you just take the output from the last step after removal of variables with $P > \frac{1}{2}$. Still now clear on how a full model fit uses less memory than stepwise.
Apr
16
comment Is it better to use a random sample with 15% response rate or quota sample?
If I understand the setup, yes.
Apr
16
comment Is it better to use a random sample with 15% response rate or quota sample?
If your response rate is $< 1.0$ (i.e., if there are any non-responders) you can pretty much rest assumed that the sample is not random. A response proportion of 0.15 typically means the sample is not reliable for most purposes.
Apr
16
comment Trust of coefficients of Logistic Regression
If the size of the dataset is too big to numerically handle stepwise regression, I don't understand how it is not too big to fit the full regression model. But to answer your question there is no real need to remove any variables. If you used $\alpha=0.5$ you would do little damage, and remove a few variables though.
Apr
16
answered Trust of coefficients of Logistic Regression
Apr
14
answered How to calculate Area Under the Curve (AUC), or the c-statistic, by hand
Apr
10
comment Mood's median test - an apparent paradox
citeulike.org/user/harrelfe/article/13265551
Apr
10
comment Mood's median test - an apparent paradox
Many statisticians consider this test to be obsolete. That was the subject of an article in The American Statistician some years ago. It is too inefficient, and it is easy to see why because of the information it discards.
Apr
9
comment Minimizing symmetric mean absolute percentage error (SMAPE)
Bad nomenclature leads to many problems downstream. But more importantly make sure the index is one you really want to optimize.
Apr
9
comment Minimizing symmetric mean absolute percentage error (SMAPE)
Since there is no 100 in the formula, "percentage" should be dropped.
Apr
8
answered Change scores as outcome measurements in clinical trials
Apr
8
answered Transforming continuous variable to ordinal for estimation with ordered logit
Apr
8
awarded  Notable Question
Apr
7
comment Why do we use the histogram?
Related to @NickCox's comments, I frequently prefer a spikey distribution summary, i.e., what some of my R functions call a "spike histogram" that shows up to 100 bins, and if there are < 100 unique values, all the points. If there are no ties this is essentially a rug plot. I like to see all the data in all their glory, which still allows me to see tendencies.
Apr
6
comment Why does hypothesis testing using coefficient and odds ratio give different conclusion?
It is asymmetric by definition because OR cannot be negative but it has no limit on the high end.
Apr
6
awarded  Nice Answer
Apr
6
answered Why does hypothesis testing using coefficient and odds ratio give different conclusion?