Statistics for everyone: ch.20 Paired t-test and Wilcoxon signed rank test

easier R than SPSS with Rcmdr : Contents

ch.20 Paired t-test and Wilcoxon signed rank test

Write 3 lines of script as shown above, select them all, and click ‘Submit’.

Line 1 mean to install a package called a datarium.

Line 2 means to run the datarium package.

Line 3 means to load the data named mice2 in this package.

In some cases, a window appears like the left. Then just click ‘OK’.

When you click on the red square part, it selects the mice2 data because it is ready to be active data. In other words, since the mice2 data has been loaded, but it has not yet become active data, we need to make this as active data.

This is very small data, and it weighs 10 rats before and after they have done something. Depending on the treatment, the weight may vary.

Select paired t-test and then select two numeric variables to finish.

p<0.001 is statistically significant, and the difference value is also shown as a confidence interval.

The mean of the two groups and the standard deviation of each are also represented. What should be expressed in the paper, however, is not only the mean and standard deviation of each group, but also the mean of the difference values shown just above and the confidence intervals thereof. For this data, the mean of the difference value is -199.4, and its 95% confidence interval is (-217.1, -181.8).

Now select ‘Wilcoxon signed rank test’.

It is a nonparametric test that corresponds to paired t-test, and the menu configuration is almost identical. There are 4 statistical methods, starting with Exact.

Depending on the statistical method, the results vary slightly. Now you know what Exeact and continuity correction mean, as you learned earlier in the chi-square test. The default seems to be chosen appropriately depending on the situation, in this case the same result as Exact. If the number is small, Exclude is more recommended.

If you draw it in Excel, the graph will look like this. We’ll learn how to draw graphs in the next chapter.

As you can see, the later results are much higher, which is consistent with the results of the tests from the paired t-test or the Wilcoxon signed rank test. However, we do not know if this is the effect of any treatment, whether it is because of the passage of time, or whether it is due to learning. This is an inherent limitation of before-after research. Since we take 2 measurements on one object, the two data have a deep relationship with each other. In a similar way, the same statistical method can be applied to the application of different treatments to a person’s two arms or two legs.

‘paired t-test’ is called ‘dependent t-test’, ‘paired student’s t-test’, ‘t-test for matched pairs’, ‘repeated samples t-test’, ‘related samples t-test’, etc. And the ‘Wilcoxon signed rank test’ is also called the ‘Wilcoxon t test’.

It is better not to do a test that compares before and after a treatment unless it is necessary to present a lot of it in a paper. If it’s a very early study, a treatment that has just been introduced, you might be wondering. But it’s so natural that your treatment will improve, so you don’t usually wonder about that. For older treatments, comparing existing and new treatments is usually an area that researchers and readers are curious about.