Viewing By Month : October 2004 / Main
October 29, 2004
What is the difference between ordinal, interval and ratio variables? Why should I care?
Over the years, I have heard a few people criticize my book Intuitive Biostatistics since it doesn't begin with an explanation of the different levels of measurement. So I wrote a short article explaining the differences between categorical, ordinal, interval and ratio variables.
October 23, 2004
Is it helpful to use ANOVA post tests to compare two dose-response curves at every dose ? Compared to using a t test, is it always harder to find 'significance' when I use a post test?October 12, 2004
How do log-rank, Mantel-Haenszel, and Gehan-Wilcoxon tests differ?
The log-rank test for comparing survival curves is equivalent to the Mantel-Haenszel method.
What is difference between Type I, Type II, and Type III errors? And what is a Type 0 error?
The Gehan-Wilcoxon method (also attributed to Breslow) gives more weight to deaths at early time points. In contrast, the log-rank test gives equal weight to all time points. Prism 4 does not perform this test, but we plan to offer this test in Prism 5. More
The terms Type I and II errors are commonly used to mean false positive and false negative statistical results. The term Type III error is used in two different ways. One humorous definition is that a Type III error is the correct answer to the wrong question. Another definition is that you've made a Type III error when you correctly conclude that a difference is statistically signicant, but have the direction of the difference backwards. More...
If one-way ANOVA results are not significant, should I still look at post tests?
Yes. Don't make the common mistake of skipping over the post test results unless the overall ANOVA has a P value less than 0.05.
