# Nonparametric Statistical Analysis In Comparative Psychology Research

## Submission Type

Event

## Faculty Advisor

Ellen Furlong

## Expected Graduation Date

2018

## Location

Center for Natural Sciences, Illinois Wesleyan University

## Start Date

4-21-2018 2:00 PM

## End Date

4-21-2018 3:00 PM

## Disciplines

Education

## Abstract

Parametric statistics, the analytical tools most commonly used in experimental psychology, rely on a number of statistical assumptions that are not always met in psychology research. One such assumption, the assumption of normality, demands that experimental data is normally distributed such that the majority of the data clusters around the mean. Many psychology experiments involving human populations do produce normally distributed data; however, data collected from experiments involving nonhuman animals are rarely normally distributed. Small sample sizes, the use of categorical or ordinal variables, and elevated variance represent some of the many factors contributing to the non-normality in comparative psychology data sets. In such cases, an alternate set of analytical tools, nonparametric statistics, enable researchers to more accurately analyze data as these tests do not rely on the same sets of assumptions as typical parametric tests. Using these nonparametric statistical tests we reanalyzed data from published research that originally utilized parametric statistics when nonparametric tests would have been more appropriate. Nonparametric analyses revealed different results from the parametric tests reported in several notable experiments suggesting that the conclusions were subject to alternative interpretations. We suggest that researchers should become aware of the assumptions of parametric statistics and be vigilant in selecting appropriate statistical tests. Comparative psychologists in particular may benefit from adding nonparametric statistical tests to their analytical tool sets.

Nonparametric Statistical Analysis In Comparative Psychology Research

Center for Natural Sciences, Illinois Wesleyan University

Parametric statistics, the analytical tools most commonly used in experimental psychology, rely on a number of statistical assumptions that are not always met in psychology research. One such assumption, the assumption of normality, demands that experimental data is normally distributed such that the majority of the data clusters around the mean. Many psychology experiments involving human populations do produce normally distributed data; however, data collected from experiments involving nonhuman animals are rarely normally distributed. Small sample sizes, the use of categorical or ordinal variables, and elevated variance represent some of the many factors contributing to the non-normality in comparative psychology data sets. In such cases, an alternate set of analytical tools, nonparametric statistics, enable researchers to more accurately analyze data as these tests do not rely on the same sets of assumptions as typical parametric tests. Using these nonparametric statistical tests we reanalyzed data from published research that originally utilized parametric statistics when nonparametric tests would have been more appropriate. Nonparametric analyses revealed different results from the parametric tests reported in several notable experiments suggesting that the conclusions were subject to alternative interpretations. We suggest that researchers should become aware of the assumptions of parametric statistics and be vigilant in selecting appropriate statistical tests. Comparative psychologists in particular may benefit from adding nonparametric statistical tests to their analytical tool sets.