Power Analysis is a built-in tool that calculates how large of a sample you need to correctly detect a statistically significant result. For example, you enter your group means, standard deviations, desired significance level, and a target power level, and Prism tells you the minimum sample size required.
This tutorial covers how to run a Power Analysis in Prism Cloud for unpaired t tests, paired t tests, and ANOVAs, and how to adjust significance thresholds and effect size to fit your study design.
Where to find Power Analysis in Prism: The Power Analysis tool is accessible from the Prism home screen. Selecting "Sample Size and Power" and “Start Analysis” opens the tool in Prism Cloud, where all calculations run.
How to calculate sample size for a t test in Prism: Enter group means and standard deviations, set your target power, and Prism calculates the required sample size per group. At 80% power, detecting a small but real difference between groups may require as few as 39 samples per group.
How power, effect size, and sample size relate to each other: Increasing your required power increases your required sample size. Prism displays this relationship as a live chart and table so you can see exactly what tradeoffs you are making.
What happens when you raise your significance threshold: Raising the required power from 80% to 99% can more than double the sample size needed, even when the effect size stays the same.
How to run Power Analysis for ANOVA in Prism: For experiments with three or more groups, Prism accepts group means and standard deviations, between-group variance and within-group variance, or effect size to calculate the sample size needed across all groups.
How do I calculate sample size in Prism?
Prism calculates sample size through its Power Analysis tool, found on the home screen under "Sample Size and Power" and “Start Analysis.” You select your test type, such as an unpaired t test or ANOVA, then enter your expected group means, standard deviations, and target power level. Prism returns the minimum sample size per group needed to achieve that power.
What is statistical power and what does 80% power mean?
Statistical power is the probability that a study will detect a true difference between groups when one actually exists. A power of 80% means that if the groups truly differ, your study has an 80% chance of returning a statistically significant result and a 20% chance of missing the effect entirely. A power of 80% is a common standard in research, though some studies require higher thresholds.
Do I need existing data to run a Power Analysis in Prism?
No. Prism lets you run a power analysis using either real data or estimated values. If you have pilot data, you can enter your actual group means and standard deviations. If you are in early planning, you can enter an estimated effect size instead and Prism will calculate sample size based on that assumption.
Should I use a one-tailed or two-tailed test in Prism?
In most cases a two-tailed test is the right choice. A two-tailed test accounts for the possibility that either group could end up higher or lower than the other. A one-tailed test assumes the difference can only go in one direction and should only be used when there is strong prior evidence supporting that assumption.
How does effect size affect sample size in a Power Analysis?
Effect size measures the strength of a relationship between variables. For example, effect size for a t test is defined as the separation between group means in units of standard deviations. A larger effect size means the groups are more distinct, so fewer samples are needed to detect the difference. A smaller effect size requires a larger sample size to achieve the same level of statistical power. In Prism, adjusting the effect size updates the required sample size in real time.
How do I know what effect size to choose?
Effect sizes are typically classified as small, medium, or large. What value this corresponds to depends on the type of test you are planning. For example, an unpaired t test might use Cohen’s d as a measure of effect size. The general rule of thumb for this value is: small (0.2), medium (0.5), and large (0.8). You should pick an effect size appropriate for your experiment based on historical data or professional judgement.
Is Power Analysis available in Prism Cloud?
Yes. Power Analysis in Prism runs through Prism Cloud. When you select the power analysis option from the Prism home screen, the tool opens online and connects to your Prism Cloud account. No separate installation is required.
