Take the 2-minute tour ×
Cross Validated is a question and answer site for people interested in statistics, machine learning, data analysis, data mining, and data visualization. It's 100% free, no registration required.

Suppose I have $n=100$ observations of ordinal data and get threshold coefficients $b_1, \dots, b_3$ and probit slope $b_4$. I want to test the hypothesis $H_{0}: \frac{b_{3}}{b_{4}} = \frac{1}{2}$ vs. $H_a: \frac{b_{3}}{b_{4}} \neq \frac{1}{2}$. So I want to count the proportion of times we fail to reject the null hypothesis. To do this, would we just keep sampling 100 observations similar to the previous ones and return the proportion of p-values greater than $0.05$?

share|improve this question
2  
To do power analysis you need to specify the effect size - if $\frac{b_3}{b_4} = \frac{1}{2} + 10^{-20}$ you will have much lower power than if $\frac{b_3}{b_4} = 5000$. So, you need to decide what is a "small" and "large" effect size in the context of the application, simulate data under those circumstances, then empirically look at how often you reject $H_0$. (btw, I think you meant proportion of $p$-values less than $.05$:)) –  Macro Jan 17 '13 at 16:27

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Browse other questions tagged or ask your own question.