For example, we could have assumed that mean NDI was linear, with possibly different slopes/intercepts for the Treat/Sham groups. Note: we could have assumed a simpler model for the means. It was known from other studies that NDI scores have a standard deviation of around 12, and those have been observed to decrease over time. The details we need include: A) prior knowledge of how average pain decreases for people in the Sham group, B) some idea about the variability of scores, C) how scores would be correlated with one another over time, and D) how much better the Treat group would need to be in order for the new procedure to be considered clinically meaningful.Īs a first step, the PT sat down and filled in the following table.Īll of the entries in the above table represent population mean NDI scores for people in the respective groups at the respective measurement times, and were filled in based on prior research and educated guesses by the PT. Of course, this information had been hiding behind the scenes all along, even with those old research papers and online calculators, but the other methods make it easy to gloss over the details, or they’re so complicated that researchers will give up and fall back to something like Cohen’s T-shirt effect sizes. What you find when you start down this path is that there is a lot of information required to be able to answer the question. The value we get is just an estimate of the power, but we can increase the precision of our estimate by increasing the number of repetitions in step 3. If the power isn’t high enough, then increase the given sample size and start over. An estimate of the power (for that sample size) is the proportion of times that the test rejected.
Using gpower to calculate sample size for manova how to#
So my buddy and I worked it all out from start to finish for this simple example, in the hopes that by sharing this information people can get a better idea of how to do it the right way, the first time. Nowadays the fashionable people say, “Just run a simulation and estimate the power,” but the available materials online are scantly detailed. Russell Lenth calls these “T-shirt effect sizes”. Some of the calculators advertise Cohen’s effect sizes, which are usually stated something like “small”, “medium”, and “large”, with accompanying numerical values. And I was never really sure whether I’d got it right, or if I had screwed up with a parameter somewhere. Then I resorted to online calculators (the proprietary versions were too expensive for my department!), which are fine, but they all use different notation and it takes a lot of time poring through documentation (which is often poor, pardon the pun) to recall how it works. I used to go look for research papers where somebody’s worked out the F-test and sample sizes required, and pore over tables and tables. I encounter some variant of this question a lot. “On what?” Well, it depends on how good the statistical test is, and how good your method is, but more to the point, it depends on the effect size, that is, how far apart the two groups are, given that the method actually works. Now comes the question: how many people should you recruit for your study? The answer is: it depends. There is some brief information about the index here. The index ranges from 0 to 100 (more on this later). Pain is measured by an index (there are several) the one we’re using is something called NDI, which stands for “Neck Disability Index”. For each Subject, you measure their pain at time 0, 15 minutes, 48 hours, and 96 hours. (Yes, those guys should really be using mixed-effects models, but those haven’t quite taken off yet.) There are two groups: the “Treatment” group does your new exercise method, and a “Sham” group does nothing (or just the placebo exercise method). The standard approach in the PT literature to analyze said data is repeated measures ANOVA.
How can you demonstrate that your method is effective? Of course, you collect data and show that people using your method have significantly lower neck pain than those from a control group. Suppose you are a PT, and you’ve come up with a brand new exercise method that you think will decrease neck pain, say.
0 kommentar(er)
