Increase alphaConduct a one-tailed testIncrease the impact sizeDecrease random errorIncrease sample size

Sound for this reason simple, right? The truth is that although these 5 ways all work-related theoretically, you could have trouble v some in practice.

You are watching: Determine two ways power can be increased

1. Alpha is quite well collection at .05 for most scientific studies (there are rare exceptions). For this reason you’re not going to get away v this one.

2. One-tailed tests have nothing wrong v them theoretically. However, there are just a couple of tests in which they’re even possible–namely t- and also z-tests. For this reason they’re not supplied much and have therefore show up dubious. Many reviewers won’t think you that you really were hypothesizing that direction (even if it’s obvious). They will assume you’re trying to artificially obtain that p-value lower (it has been done).

So, as soon as again, uneven you’re in one enlightened field, or one in which one sided tests are generally done, you have the right to forget this one too.

3. Uneven you’re coming up v a an ext precise means to measure her constructs, it’s most likely that the result size is a large as it’s going to get. Save going.

4. Aha, miscellaneous we can work with. There are two great ways to alleviate random error. One is to do it not random. Define it with a manage variable, turning into defined variation.

See more: When Does The Elf On The Shelf Come Back 2019, When Does Elf On The Shelf Start

The various other related way is to usage some sort of repeated measures design. Because we have multiple dimensions on a subject, we have the right to now different the error variance indigenous the topic variance.

5. Finally, the crux that the matter. If #4 doesn’t work, and also it won’t always, your just option is to increase sample size. (But girlfriend knew the one, right?)

Jeff says

January 17, 2021 at 10:50 am

It is my expertise that you have the right to sometimes rise power by convert to more advanced statistics testing. For example, is that true that when you use a general linear combined model (versus rm-anova) you deserve to increase power…I assume by rise sample size using long data formatting (versus wide format).

Can you suggest to some webinar or article or training which would address this? I’m additionally interested in a step-by-step explanation around how to determine sample size using GLMM (possibly through samplesizeshop).