Statistical significance testing
Paired t-test or Wilcoxon signed rank test on numeric data
Undertake a paired t-test or Wilcoxon signed rank test to test for differences between means (or medians) of paired numeric values.
- the type of test desired (paired t-test or Wilcoxon signed rank);
- desired level of confidence for confidence interval estimation;
- the desired number of digits after the decimal point in summary results; and
- two columns of data (including a header row with labels for each column). The data must be paired in rows (ie each row is a pair of corresponding measurements) and the columns must be the same length.
Note missing values in the data will be omitted from calculations.
For this analysis, the values in the first column are subtracted from the corresponding values in the second column and a t-test (or signed rank test) undertaken to test for deviation of the mean (median) difference from zero.
- a summary of statistical test results for the selected test, including 2-tailed and 1-tailed for both tails;
- Shapiro-Wilk test for normality on both sets of measurements and the differences;
- Note: for groups with < 4 or > 5000 observations it is not possible to calculate Shapiro-Wilks W statistic to test for normality of distribution;
- a numeric summary of the data for each measurement and the difference, including histograms and normal plots;
- Note: for groups with only one observation or with standard deviation = 0 histograms and quantile plots are not shown; and
- boxplots of the data by group and a plot of confidence intervals about the mean (or median) value for each group.