Introduction
Parametric and non-parametric statistical tests are used to analyze data. The main difference is whether the test made assumes the data follow a particular distribution usually known as normal distribution or not.
Definition
1). Parametric tests
Parametric tests are statistical tests used to analyze numerical data and also make assumptions about the population on data that has been collected.
An example: Suppose a researcher wants to know whether a new drug lowers blood pressure more effectively than an old drug.
The blood pressure readings (121,128,135,140 mmHg). These are continuous numerical data. The readings are normally distributed by the researcher through a parametric test known as the independent test to compare the average pressure between the two groups.
2). Non-parametric tests
Non-parametric tests are statistical tests that do not require to follow a specific distribution such as normal distribution. They are used when assumptions for parametric tests are not met.
Example: A doctor compares patients' pain ratings before and after receiving physiotherapy.
Differences between parametric and non-parametric tests.
1). Parametric tests compare means (averages) whereas non-parametric tests compare median, ranks, or frequencies.
2). Parametric tests are more powerful when assumptions are met whereas non-parametric tests are powerful but more robust when assumptions are violated.
3). Parametric tests are more sensitive to extreme outliers whereas non-parametric are less sensitive to outliers because many tests use rank instead raw values.
4). Parametric tests often perform best with moderate or large samples whereas non-parametric are well suited with small samples.
5). Parametric tests assume that the data follows a specific distribution (normal) whereas non-parametric tests do not require the data to follow a specific distribution.
Use of parametric and non-parametric tests and their role in data science.
1). Use of Parametric Tests in Data Science
a). Hypothesis testing.
b). Predictive analytics.
c). Measuring relationships among variables.
d). Evaluating machine learning models using statical comparisons.
2). Uses of Non-Parametric Tests in Data Science
a). Analyzing survey and questionnaire data.
b). Working with non-normal or incomplete datasets.
c). Detecting associations in categorical variables.
d). Validating results when parameters assumptions are not met.
In conclusion parametric and non-parametric tests enable data scientists to choose appropriate analytical methods for different types of data leading to more reliable concluding and better-informed decisions.
Top comments (0)