The paired comparisons test dialog is shown below. Here, you enter the information that Maritz Stats needs to compute the results of a paired comparisons test.
Show me how to use it! To see an example of how to use the paired comparisons test, click here.
Show me how it works! For more information on the formula used for the paired comparisons test, click here.
The Base is the total number of respondents, including No Preference and No Answer. If you enter percentages, this number will be used to calculate the raw number of respondents in each category.
For each of the two groups you are comparing, you can enter either percentages or raw number who fell into that group. Selecting either Percent Selecting or Number Selecting enables the appropriate controls to enter this information.
When conducting a paired comparisons test, you can make one of two assumptions. Either those respondents who indicate no preference can be ignored, or you can assume that this percentage/number will be evenly split between the two categories. If you choose to split these responses, enable this check box, and then enter either the number or percent of respondents who fell into this category.
If you are interested in detecting significant differences in one direction (for example, preference for A is GREATER THAN preference for B), then select a one-tailed test. If you are interested in detecting significant differences in either direction, then select a two-tailed test.
The confidence level indicates the degree of confidence that the interval actually encompasses the unknown population parameter variable. For most applications, the default of 95% should be used. However, you have the option of being more stringent (using the 99% confidence level) or less stringent (90% or 80% confidence level) based on your needs. The higher the confidence level, the less likely it is that differences will be found significant.
The results are presented in three stages. First, the z-value is presented. This can be used to compare to a z-table if a significance level other than 80%, 90%, 95%, or 99% is desired. Second, the critical value for this test is presented based on the significance level selected. Finally, a simple "yes" or "no" indicates if these two groups are significantly different based on the information provided.