Overview of the app and user instructions

About the app

The aim of the app is to give students and researchers a tool to balance experimental groups before starting an experiment. Prior to an experiment, researchers can use the app to balance groups both in terms of measured variables and possible confounding covariates. The app shows comparisons between different levels of a factor, and also the interaction results of all the levels of two independent factors.
The app provides the main effects of the analysis, as well as post hoc test results. The comparisons between experimental groups are also presented graphically. Two sets of figures are provided: firstly, for Frequentist analysis and secondly for Bayesian analysis. Post hoc analyses are presented simultaneously in Frequentist and Bayesian paradigms to give the oppurtunity to compare the different approaches.

User instructions

Settings

  • The “Select dataset to use” option allows you to choose one of the provided datasets or your own uploaded dataset.
  • The “Browse” button provides the opportunity to upload and analyze your own data. You can upload several datasets, and you can switch between the datasets with the “Select dataset to use” button. The data needs to be in .csv format. Press “Add uploaded data to selection” button to add your data to the avaliable datasets. Prior to adding your data to the collection you can name it in the “Give name for dataset” box.
  • The “Select parametric post hoc to use:” option enables you to choose from three types of post hoc tests for Frequentist analysis.
  • With the “p-value” and “Bayes factor” sliders you can specify the significance level for Frequentist analysis and the Bayes Factor value for Bayesian hypothesis testing. Those cut-off values are used to indicate significant differences between experimental groups. The significant differences are displayed as red lines in the figure.
  • After selecting the dataset, refresh the variable selection by pressing the “After data upload - update variable selection” button.
  • In the “Select dependent variable: ” box you can choose the dependent variable for your analysis. You can choose among variables that are encoded as numerical in the data.
  • In the “Select independent variables: ” box you can choose one or two independent variables for your analysis. These are the factors you want to compare in terms of the dependent variable. The factors can have from 2 to 8 levels. You can choose among variables that are encoded as characters in the data.
  • To calculate results press the “Compute” button. Navigate between the output tabs to see the results of the analysis.
  • The “Adjust plot height:” slider allows you to adjust the heights of the post hoc plots in the output.

Output

  • The “Explore the data” tab enables taking a closer look at the selected data.
  • The “Main Effects” tab gives an overview of the main effects of the analysis.
  • The “Post hocs in table” tab gives an overview of the post hoc tests. This analysis is presented simultaneously in Frequentist and Bayesian paradigms. Intervals of Frequentist +/- 1 SE and corresponding Bayesian 68% HDI are denoted on graphs as error bars.
  • The “Post hocs plotted on a graph” tab shows the post hoc test results graphically.

Authors

Mait Metelitsa mait.metelitsa@gmail.com
Martin Kolnes martin.kolnes@ut.ee
University of Tartu