## Help With Conducting a Power Analysis

# Conducting a Power Analysis

**Conducting a power analysis** is challenging for most students. Power analysis is done using various software products, some of which are easy to use. The most popular software used is GPower due to its numerous benefits. It is user-friendly and can be downloaded for free for both personal computers and Mac. The product also supports various designs, including t-tests, ANCOVA, correlations, ANOVA, proportions, regression, chi-square, repeated measures, non-parametric equivalents, and logistic. Additionally, it has an effect size calculator, and a tutorial manual can be found online to train new users on its use.

## Steps to Follow When Conducting a Power Analysis

- The researcher needs to select the type of power analysis they will do, whether it is priori, post-hoc, criterion, or sensitivity.
- They should choose the expected research design such as t-test or ANOVA that reflects their hypotheses of interest.
- The researcher should select a power analysis tool that supports their study design.
- Give three of the four parameters, usually power=.80, alpha=.05, expected effect size supported by previous literature or pilot data.
- Solve for the unknown remaining parameter, the sample size.

## Reporting Power Analysis Results

Researchers usually conduct a power analysis for the central hypotheses of interest. In complex research plans, the researcher may be forced to perform separate power analyses for each hypothesis and choose the larger sample size as a basis of recruitment. Attrition and control for possible moderators are other considerations in conducting and reporting power analysis to estimate sample size. You should consider the expected moderators when choosing the research design in the power analysis tool. Although every research varies, writing a power analysis should not be a complex process. It is advisable to have one or two paragraphs outlining your research plans and address prospective contingencies for more variables and attrition. Mist funders are usually interested in the feasibility of the study and a thorough study plan, which comprises a power analysis.

## Running Power Analysis

As the study progresses, a researcher may want to check the effect size to see if the expected effect is realistic in their research with its unique purpose and design. If the effect is too high or too low, the investigator will have to adjust recruitment accordingly and retain adequate power to demonstrate the effect inferentially. This can be done by running a power analysis. It is ethical to run a power analysis if it allows the researcher to prove a practical effect inferentially, and it can be included in the literature as a statistically significant and essential finding. It can be a waste of useful resources and time if the researcher wants to increase the sample size using running power analysis. A running power analysis supports a valuable effect with inferential statistics. A beneficial effect will be specific to the field of research and needs to consider risk, benefit, and cost. In G*power, you should indicate your interest in performing a priori power analysis and then select one tail t-test between two independent subjects with the same group sizes. Next, you should enter the expected effect size from pilot studies of 1.o, or you could let G*power calculate the effect size from the pilot data given, then enter the power of .80 and an alpha of .05 and press calculate. G*power will display on the same monitor the sample size needed to detect the effect level with p-values using a t-test.