XLSTAT: Two Sample T-test

 

How to do a 2 sample T-test

Please follow this procedure below to launch two-sample t-tests:

• Click on the Insert Function button to access our different functions.

• Select XLSTAT in the Or select a category field.

• Select the XLSTAT_TTest function to run a Student t test in order to compare two paired samples.

• Click on the cell where you want to insert the function and obtain its result. For example, select the A2 cell.

• Select the XLSTAT_TTest function and click OK.

• Select the two data ranges, excluding sample labels to avoid errors.

• Fill only the first two fields. Enter TRUE in the Pairedfield, as the data are paired.

• The last option, PValOut, is true by default to display the p-value. Setting it to false shows the Student’s t statistic instead.

• Click OK to run the t-test and see the results in Excel.

 
The p-value is 0.36 which means that the null hypothesis that the samples have an identical mean cannot be rejected with a significance level of 0.05.

 

Which are the Arguments for a 2-sample T-test?

The arguments of the TTest function are:

• V1: range for the first sample

• V2: range for the second sample

• Tail: 0 for a two-sided test, -1 for a lower one-sided test, 1 for an upper one-sided test

• HypDiff: the hypothesized difference between the two samples

• Paired: TRUE if the two samples are paired

• EqualVar: TRUE if the two variances are assumed to be equal

• pvalOut: TRUE (default) if you want to display the p-value, FALSE if you want to display the t statistic.

The Student’s t-test assumes the samples come from normally distributed variables. If this is doubtful, use the Wilcoxon signed-rank test, a nonparametric alternative that doesn’t require this assumption.

What are the Arguments Used for a Wilcoxon Signed Rank Test?

The arguments of the WilcoxonTest function are:

• V1: range for the first sample

• V2: range for the second sample

• Tail: 0 for a two-sided test, -1 for a lower one-sided test, 1 for an upper one-sided test

• Correction: TRUE if the correction should be applied with the approximation (default is TRUE)

• ExactTest TRUE if you want to run the exact test (default is TRUE)

• pvalOut: TRUE (default) if you want to display the p-value, FALSE if you want to display the V statistic.

The Wilcoxon signed-rank test yields a p-value of 0.43, supporting the conclusion that the samples do not differ in location.

For unpaired data, use XLSTAT_TTest with Paired = FALSE for a t-test if data are normal, or the Mann-Whitney test for a nonparametric alternative.