15,435 reputation
12262
bio website biostat.mc.vanderbilt.edu/…
location
age
visits member for 3 years
seen 4 hours ago

9h
comment Should I be using a Welch or standard T-test?
It is often better to generalize the question to ask if a randomly selected observation from A tends to be larger than a randomly selected observation from B. This is the essence of the highly efficient and robust Wilcoxon-Mann-Whitney two-sample rank test.
20h
answered Adjusting significance level with for different kinds of tests (chisq, mann whitney etc.) on the same variable
2d
comment How can I improve the predictive power of this logistic regression model?
Any method that dichotomizes a continuous variable (in this case the logistic model's predicted probability) is highly problematic. Now couple that with the fact that the threshold is completely arbitrary and the proportion classified correctly is an improper scoring rule (i.e., it is optimized by a bogus model) you have a perfect storm.
2d
comment How can I improve the predictive power of this logistic regression model?
This was never a sound idea. I very much regret including a classification table as an example in the users guide for the first SAS procedure for logistic regression. In my view classification tables represent bad statistical practice.
2d
answered How can I improve the predictive power of this logistic regression model?
2d
comment Tree structured proportion of success : statistical significance?
Sorry I misunderstood. But "cutting the nodes where the proportions of success are not statistically significant" needs clarification then. Exactly what is the hypothesis being tested, if you are interested in "significance"?
2d
answered Tree structured proportion of success : statistical significance?
Apr
15
comment I want to learn about ROC curve — what is the canonical textbook?
The first thing is to learn when ROC curves are relevant/helpful. I see them being used when it would be much better to use risk prediction. They are not useful for determining a cutoff when a cutoff doesn't exist or when you are seeking a cutoff on an independent variable. In general, ROC curves are at odds with individual decisionmaking.
Apr
15
comment Treating ordinal variables as continuous for regression problems
@Scortchi I did not know about that R package. Wonderful. tomka if it is $Y$ that is ordinal then the proportional odds, proportional hazards, or other ordinal models may indeed be excellent choices.
Apr
14
answered Treating ordinal variables as continuous for regression problems
Apr
14
revised Treating ordinal variables as continuous for regression problems
edited tags
Apr
14
comment Unequal variances t-test or U Mann-Whitney test?
I don't think that is quite accurate. It's probably the case that the bias is not huge if you don't have an eye to the test result, but a bias can nontheless creep in, in a subtle way. Simulations will answer the question.
Apr
13
comment Unequal variances t-test or U Mann-Whitney test?
Nonparametric independent group comparisons (unpaired data) are invariant to transformation of $Y$. For methods that require transformations to be "right" (usually, parametric methods) the act of finding the transformations creates tremendous uncertainty that can greatly effect the true coverage of the "final" confidence interval and can affect type I error. This is described in an excellent paper by J Faraway: The Cost of Data Analysis, J Comp Graph Stat 1:213, 1992.
Apr
13
comment Unequal variances t-test or U Mann-Whitney test?
Multi-step = pre-testing (equal variances, normality, etc). You're right that failure to satisfy model assumptions will result in worse type II error, but if you entertain non-normality or non-equal variances it is better to just allow for these up front and use a nonparametric test.
Apr
12
comment Why is logistic regression a linear classifier?
Well said. Thank goodness logistic regression is not a classifier.
Apr
12
comment Unequal variances t-test or U Mann-Whitney test?
That multi-step procedure will not preserve the type I error of the test.
Apr
12
comment Unequal variances t-test or U Mann-Whitney test?
It is not correct to say that the $t$-test is generally more powerful. The Wilcoxon test has $\frac{3}{\pi}$ efficiency under normality and is usually much more efficient under non-normality. And in line with another earlier comment, it is not good practice to call statistical tests by their SPSS names.
Apr
10
awarded  Nice Answer
Apr
10
comment How to further reduce predictive error in a regression tree model
Your sample size is too low by perhaps a factor of 100 for a single tree to reliably represent the predictive patterns. Your sample size is also woefully inadequate for split sample validation to be reliable and stable.
Apr
10
awarded  Nice Answer