Welcome to the Polarization Analysis App
This Shiny app accompanies the preprint titled 'The many flavors of polarization: A systematic comparison of different conceptualizations and contexts' by Olivia Fischer and Renato Frey. Explore the app to learn more about polarization in survey response data.


Option 1

Understand how different distributional shapes relate to operatioalizations of polarization.

Option 2

Examine the convergent validity of different operationalizations of polarization in your own or various other data.





The R code for this app is available on OSF . Please cite the app via the preprint: https://doi.org/10.31234/osf.io/bv496 .

For details on the operationalizations as well as their equations, please see the preprint: https://doi.org/10.31234/osf.io/bv496

Range: The range or spread of the distribution is simply the difference between the highest and lowest value. The higher the range, the more polarized the underlying distribution

Median absolute deviation from the mean (MAD): The MAD is a measure of a distribution's variability. The larger the MAD, the more polarized the underlying distribution.

Relative size difference: Relative size difference measures the difference in the relative number of people who place themselves/their opinion below vs. above the scale midpoint. The larger the size difference, the less polarized the distribution.

Relative distance: Relative distance refers to the difference in mean attitudes of people below vs. above the scale midpoint. The larger the distance, the more polarized the distribution.

Polarization index: The polarization index is a multiplicative combination of relative size difference and relative distance. The larger the index, the more polarized the distribution.

Bimodality coefficient (BC): The BC uses a distribution's skewness, kurtosis, and sample size to estimate how bimodal a distribution is. A BC of .56 corresponds to a uniform distribution. A BC above .56 indicates polarization.

Ranks: Lower ranks (i.e., 1, 2, 3) and darker reds indicate a higher degree of polarization according to the respective operationalization. Higher ranks and lighter shades of red indicate a lower degree of polarization.

High convergent validity: We can speak of high convergent validity or high agreement between operationalizations of polarization when the ranks or colors for individual items (i.e., rows) across multiple operationalizations (i.e., columns) are similar. This generally corresponds to a higher mean absolute rank correlation in the correlation plot.

Low convergent validity: We speak of low convergent validity or low agreement between operationalizations of polarization when the ranks or colors for individual items (i.e., rows) across multiple operationalizations (i.e., columns) are dissimilar. This generally corresponds to a lower mean absolute rank correlation in the correlation plot.