ANOVA_power

Using this App

This Shiny app is for performing Monte Carlo simuations of factorial experimental designs in order to estimate power for an ANOVA and follow-up pairwise comparisons. This app allows you to violate the assumptions of homoscedascity and sphecity (for repeated measures). Also, the simulations take a considerable amount of time to run. If you don't need/want to violate these assumptions please use the ANOVA_exact app.

Click here for the other app

The Design Tab

You must start with the Design tab in order to perform a power analysis. At this stage you must establish the parameters of the design (sample size, standard deviation, etc). Once you click Submit Design the design details will appear and you can continue onto the power analysis.

Power Simulation Tab

In this tab, you will setup the Monte Carlo simulation. You will have to specify a correction for multiple comparisons (default=none) and the alpha level (default=.05). If you have repeated measures you will need to specify the sphericity correction (default=none).

Download your Simulation

Once your simulation is completed a button a button will appear on the sidebar to download a PDF

NEWS

Current updates to Superpower's Power Shiny App

Option for estimated marginal means added

Now allows for unequal n input

Inputs

Specify the factorial design below
*Must be specficied to continue*

Add numbers that specify the number of levels in the factors (e.g., 2 for a factor with 2 levels). Add a 'w' after the number for within factors, or 'b' for between factors. Seperate factors with an asterisks. Thus '2b*3w' is a design with two factors, the first of which has 2 between levels, and the second of which has 3 within levels.

Design Input

Would you like to enter factor and level names?

Specify one word for each factor (e.g., AGE and SPEED) and the level of each factor (e.g., old and yound for a factor age with 2 levels).

Factor & level labels

Would you like to enter different sample sizes per cell?

Sample Size per Cell

Would you like to enter multiple standard deviations? *Warning: Violates homoscedascity assumption*

Specify the list of standard deviations.

Common Standard Deviation

Would you like to enter a correlation matrix (rather than a single correlation)? *Warning: may violate sphericity assumption*

Common correlation among within-subjects factors

Specify the correlation matrix. If a correlation is entered across between subjects factor it will be reset to zero. Values only need to be entered in the upper triangular part and the app will take it from there.

Note that for each cell in the design, a mean must be provided. Thus, for a '2b*3w' design, 6 means need to be entered. Means need to be entered in the correct order. The app provides a plot so you can check if you entered means correctly.

Means for Each Cell in the Design

Click the button below to set up the design - Check the output to see if the design is as you intended, then you can run the simulation on the next tab.

Design Output

WARNING: Ensure sample sizes are the same between levels of within subjects factors (Otherwise the app will crash).

Simulate Power for Design

Simulation Parameters

Would you like to set a "seed" for reproducible simulations?

Specify a seed number.

Seed Number

Sphericity Correction

Would you like to compare the estimated marginal means?

Keeping the default settings will result in all pairwise comparisons being performed.

What model would you like to use for the estimated marginal means

What type of comparisons would you like to make?

What correction for multiple comparisons would you like to make for the estimated marginal means? Warning: Tukey and Scheffe are not appropriate when there are within-subjects factors

What comparisons would you like to make with estimated marginal means?

The addition sign ('+') will add factors for comparisons while factors after a vertical bar '|' specifies the names of predictors to condition on

For more information on setting comparisons

Select adjustment for multiple comparisons (Note: this is meant for *exploratory* ANOVAs). This will adjust the ANOVA-level and t-test (pairwise) comparison effects.

Number of Simulations

Alpha Level