Statistics for everyone: Excel and Statistics(19), F-test

Excel and Statistics, everybody should know. : Contents

F-test

Earlier, we used a different method for the ‘t-test’ depending on whether the variances were equal or not. Let’s take a look at the ‘F-test’, which can tell about variances.

With ‘example1. xlsx’, select ‘F-Test Two-Sample for Variances’

Enter two ranges in the same way.

In this case, the p-value is too large to say the variance is unequal. In general, if p is smaller than 0.05, it is often judged that the variance is not equal .

Therefore, before you perform t-test, you should check whether the variances are equal or not. On the other hand, while the F-test is meaningful as a preparation for t-test, it can also be used in research to see if the variances are different. For example, if there are two machines that cut screws with an average diameter of 10mm and it may be a good idea to check whether the variance is the same or different to test whether it remains constant or not.

Plus Alpha

Visit https://tinyurl.com/Homogeneity-Variance and upload ‘example4.csv’.

This material is placed in columns B and C of the material we saw earlier, stretched downwards, with the names groups and value. This arrangement is called a long form, and the one on the left is called a wide form.

It shows 3 plots. The left side represents the actual value as a box plot. It also shows which groups are up or down and gives you an idea of how far they are spread. The center plot is centered on the median, allowing you to focus more on the extent of the spread up and down. The plo t on the right represents the absolute value of the median and difference, further exaggerating the degree of spread.

The ‘result’ tab shows Levene’s test first, out of five tests: Levene’s test, Bartlett test, Brown-Forsyth test, Fligner-Killeen test, and F test. Levene’s Test is also used as the default test for SPSS, etc., and can be tested in groups 3 and above.

It also shows the result of the F-test, where the p-value is exactly twice the p-value shown in Excel. In Excel, it is a one-tailed test, so it is only expressed by multiplying by 2 to make it a two-tailed test, and the F value and degrees of freedom are all the same as in Excel.

The F-test, on the other hand, is a tool that allows you to look at the variance in two groups and is not available for more than three groups. Also, from the picture above, it seems that the variance of the two groups is different (subjective), but the test seems to tell us that they both have large p-values and that the variance is not different.

Here you can see the nature of the p-value. As the number of samples decreases, the p-value tends to increase, making it easier to judge that there is no difference. The larger the number of samples, the easier it is to judge that there is a difference. Therefore, you should not judge only by the value of p, but at any time, get into the habit of plotting.

On the other hand, if the p-value is greater than the standard (not significant), it does not mean that there is no difference, but rather that it cannot be said that there is no difference, which means that there is no conclusion, and it also means that no conclusion can be drawn (with this number of samples).

On the other hand, the actual variance values are 927 and 931, which look similar. Therefore, in order to check the overall distribution, it is necessary to draw a plot and check it from various angles.

Excel and Statistics, everybody should know. : Contents

=================================================