<p>View the Shiny Application: <a href="http://www.aggieerin.com/shiny/altnhst/" rel="nofollow">go here.</a></p>
<p>Notes on how to recreate our work:
- To recreate the manuscript, you will need to download the manuscript files folder and run the Rmarkdown within that folder (you can open this file with RStudio). The markdown includes the simulation code that created the files in the output folder.
- We suggest starting by searching for "library" within our document to install the proper libraries to match our markdown.
- All documents must be kept in the same folder for the markdown to run properly.
- Additionally, the manuscript PDF version requires the use of LaTeX, however, if you change line 46 from:
<code>output : papaja::apa6_pdf</code> to <code>output : papaja::apa6_word</code> you can create a word document version of our manuscript that does not require LaTeX to be installed.
- To run the interactive application to view the graphs on your own computer, you should download the manuscript folder and run the app.R within that file. That file should open with RStudio and "Run App" will appear in the top right corner of the script window. This application should run as long as the app.R file is located within the same folder as the data files. You will need to have the shiny, ggplot2, and reshape packages installed. </p>
<p>Notes on previous drafts:</p>
<li>We originally included a version of repeated measures nonparametric NHST. We dropped this analysis from our paper, as the datasets simulated did not represent differences from parametric to nonparametric NHST (i.e. by meeting assumptions, they were nearly identical). </li>
<li><strong>Nonparametric NHST - Quade's Test</strong>: Two well-known dependent samples nonparametric tests are Friedman's test and Quade's test. The Quade test was chosen over Friedman's test because Friedman's test has lower power with only three levels (Conover, 1999). Observations are first ranked within each participant (sometimes called blocks), and ties are handled by using the average rank. After ranking, ANOVA and the <em>F</em> distribution are used to calculate <em>p</em> values. Quade's test was implemented using <em>quade.test()</em> in base <em>R</em>, and <em>post hoc</em> comparisons were calculated by using the <em>posthoc.quade.test</em> function in the <em>PMCMR</em> library with a Bonferroni correction for all possible pairwise combinations. <em>p</em>-values were binned using the same rules as described above, and exact values can be found online in the overall CSV files, as well as the percent agreement in files with <em>np</em> in their file names. </li>