Nonparametric Tests: 8 Important Considerations in Using Them

Why use nonparametric tests? When are these statistical tests used? This article aims to answer these questions.

As a student in your undergraduate years, perhaps you are more familiar with parametric tests than nonparametric tests. Whatever the reason for this tendency among statistics teachers to focus their pedagogy on parametric tests in college, you should be familiar with nonparametric tests as an alternative to parametric tests as a graduate student. Situations arise wherein you have no better option but to apply nonparametric statistics.

Why use nonparametric tests?

The use of nonparametric tests boils down to the type of data that you collect in your study. Unlike parametric tests, nonparametric tests accommodate data that have a wide range of variance. You can use nonparametric tests for both quantitative and qualitative data. These tests operate under few assumptions about the population, unlike the parametric tests that favor quantitative data analysis.

Nonparametric methods are considered robust, meaning, these tests are efficient in analyzing data characterized by outliers. Outliers are those data that deviate from the expected “normal” range of values. Unless the effects of outliers are controlled in parametric tests, they lead to erroneous conclusions. Nonparametric tests are also robust as analysis need not require data that approximate a normal distribution–more on this in the next section.

8 Important Considerations in Using Nonparametric Tests

Non-Normal Distribution of the Samples

Nonparametric tests are typically used when assumptions for normality of a sampling distribution could not be made. Non-normal sampling distributions typically results when random sampling is not possible. Bias or tendency to favor the selection of samples in one section of the population occurs.

When random sampling is not possible, skewness results in the sample distribution. Skewness refers to data’s tendency to get distributed on one side of the range of values obtained for a variable.

nonparametric test
A histogram is an effective way to show the skewness of a distribution (Image by Peggy und Marco Lachmann-Anke from Pixabay).

The skewness of a sample distribution makes parametric tests less powerful in testing hypotheses. It will be impractical to conclude using the mean as representative of a population when data is distributed almost equally, at the extremes, or clumped in the middle of the distribution. The mean will either average the low or the high values. Thus, it is not effective as a representative of the population because it is not centrally located. The mean is no longer the population’s best measure because extreme values pull the mean away from the center. Nonparametric tests work better in skewed distributions because they use the median instead of the mean.

Constraints in Data Gathering

Limiting factors in data collection that make parametric statistics impractical to use include the following: difficulty in obtaining a full list of the entire population; logistical, time, and financial constraints; unfamiliarity with the extent of the population studied; application of non-random selection of respondents or samples; tendency to commit bias in some circumstances, among others.

When Sample Size is Small

Small sample sizes tend to approximate a non-normal distribution. If you use a histogram, you would confirm this tendency as the data will show clustering on one side of the graph. You cannot account for all the possible data variations where all data points are well represented. Thus, the application of nonparametric tests is the only suitable option.

Data is Ordinal or Nominal

Parametric tests work only with continuous data such as those variables measured to obtain ratio and interval data. On the other hand, nonparametric tests can be applied to other data types, such as ordinal or nominal data.

Previous Studies Use Nonparametric Tests

In your review of related literature in your field, it helps to know how the variables you have identified in your study are measured. If those variables were analyzed using nonparametric tests, then it will be safe to do the same.

Exploring Research Topic Potential

For practical reasons, graduate researchers usually find themselves dealing with exploratory research topics. Thus, it would be best to explore a few samples to see if the issue under scrutiny is worth investigating.

The Median is the Rational Representative of Your Study

There are variables where the median represents the population instead of the mean. A society’s income, for example, is better measured using the median as very rich members of the population can significantly increase the overall mean. Getting the median distributes the range of income from either side of a distribution: 50% above and below it.

When Outliers Cannot be Removed

Outliers, which may be obtained as aberrant value from the expected values of a sample, can significantly change the mean of the sample. An outlier is a data point that deviates from the norm.


Corporate Finance Institute (n.d.). Nonparametric tests-overview, reasons
to use, types. (Accessed on 10/11/2020).

Frost, J. (2017). Nonparametric tests vs. parametric tests statistics by Jim. (Accessed on 10/11/2020).

Glen, S. (2014). Nonparametric data and tests (distribution free tests)-statistics how to. (Accessed on 10/11/2020).

Minitab Blog Editor (2015). Choosing between a non-parametric test and a
parametric test.
choosing-between-a-nonparametric-test-and-a-parametric-test. (Accessed on 10/11/2020).

Sullivan, L. (n.d.). Nonparametric tests.
html. (Accessed on 10/11/2020).

© 2020 October 11 P. A. Regoniel