Statistical Planning Regarding Confidence Intervals
Most point null hypotheses—such as H0: \( \mathsf \mu _ 1 / \mu _ 2 = 1 \) for comparing two geometric means—cannot be exactly true, yet we routinely test them anyway, and the resulting p-values are confusing and misleading, Those now shunning “p-value-ism” need sound alternatives. Bayesianism offers one route, but even its staunch advocates tolerate, and sometimes recommend, frequentist confidence intervals (Connor, 2004). How can CIs be tailored to focus on what investigators seek to know? How should sample-size analyses be performed when the focus is on CIs and not p-values?
CIplan.ttest()
Planning studies for comparing two means or geometric means using confidence intervals based on the Welch t-statistic. Not yet released.
Planning studies for comparing two means or geometric means using confidence intervals based on the Welch t-statistic. Not yet released.