Data Pilot - The significance of a correlation
The shuffling method allows us to determine the significance of a correlation between the variables measured with different scales. The attributes can be measured by the naming scale (start with “n:”), order scale (“o:”), or interval scale (start with “i:” or do not have “:”). Different attribute values are defined by the whole numbers.
As a result, you get a confidence probability value, which is calculated by subtracting the value of significance level from one.
The samples must have the same size.
If you compare the significance of a correlation of attributes with a variable, measured with the naming scale, the results can be interpreted as the attributes values for the discriminant analysis (classification with a learning sample), where the variable defines splitting of the learning sample into groups.
The results vary from 0 (no correlation) to 1 (proven correlation) and are interpreted as a confidence probability (one minus the significance level value).
If necessary, the results can be presented as a correlation matrix between the matrixes of attributes similarity.
After indicating the basic data area and the results area, select from the dropdown menu “Confidence Interval” and press the "Next >>" button.
You need to enter here the number of iterations and indicate if the correlation matrix needs to be displayed. More iterations mean higher precision level for the results and longer calculations time.
The "<< Back" button allows users to come back to the original data and calculation method selection.
Press the “Calc” button to start calculations.
During the process of calculation, a progress window appears. The “Cancel” button allows user to stop the calculations. After the calculation is complete, you get the results.