How to Compensate for a Small Sample Size in Statistical Analysis

In analytical statistics, you normally know a parameter about your population: you might know the mean or the standard deviation. You can even use a large sample size to make guesses about what those parameters might be in a population. However, if your sample size is too small (for example, you only have a sample of 10 students in a school of 2,000 students), any parameters you try to calculate won't be accurate. Instead of relying on parameters, choose a non-parametric statistical method for your data to compensate for the small sample size.

Instructions

    • 1

      Determine if your sample size is too small enough to run non-parametric tests. As a general rule of thumb, a sample size of under 100 is considered "small" by most standards. The three major non-parametric tests are Kolmogorov-Smirnov, Wilcoxon matched pairs test and the Sign test.

    • 2

      Choose a non-parametric test to run on your data. Choose the Kolmogorov-Smirnov test if you want to detect differences between means. Choose the Wilcoxon matched pairs test if you want to rank differences in observations. Otherwise, choose the Sign test,

    • 3

      Determine the importance of the study you are undertaking. An important study would be one that might risk life or health (for example, you're studying the efficacy of a new drug). If the results of your study are important, run all three non-parametric methods and compare the results.

    • 4

      Run one of the tests on your data by graphing a relevant distribution and rejecting or accepting the null hypothesis.

Learnify Hub © www.0685.com All Rights Reserved