Friday, May 18, 2012
  Search
Innovative Software for Research Reporting, online platform for survey reporting with in-depth statistical analysis for the professional expert

Survey Analysis and Reporting in the Cloud

 

Compare groups of respondents
Created by jdeboel on 10/26/2010 11:42:38 AM

The “Group Comparison” analysis is used to find out whether certain results (means) are influenced by other characteristics of the respondents (sociodemographic properties of the respondents, or his/her opinions). This is commonly described as a ‘one way anova’ procedure with pairwise comparison.

E.g. “Is there a significant difference between the way men rate the quality of our services, and the way women rate them?” But also: “Is there a significant difference between the way people who often visit our stores rate the quality of our services, and the way people who rarely visit our stores rate them?”

At the top of the page you firstly select the split variable for which you want to see if there are significant differences in the responses.  Of course, in order to do this, your dataset should have at least one question defined as a split variable. (In “Survey Setup » Questionnaire” defined as “Nominal”)

In the graphical table you’ll see your facets vertically listed with next to them the average score (Mean) they got.  In the next columns you see the results from your comparison, to see if there are significant differences. There are three possible entries in these cells:
  • A bullet (•): for this “split variable-category”; there is no significant difference with the other groups’ mean. (=The mean for all cases except this facet’s average score (mean) for this “split variable category”)
  • A horizontal bar (–): it is not possible to accurately calculate whether the cell value is significantly different or not. Most likely there aren’t enough cases for this variable in your dataset to draw a reliable inference.
  • A numerical value: there is a significant (at least 95% sure) difference with the mean value of the other groups. The value which is shown is the concept’s average score (mean) for this “split variable category”.

The calculation to decide whether there is- or is not a significant difference is done using T-test –like procedures.

If it happens that there isn’t a significant difference for a cell, but for any reason you still want to know the specific mean value, you can look at the table below the graphic. There you can see the mean values on the first line, and the standard deviation on the second line.

 <>
 An example:
A course satisfaction study (see “DEMO” dataset) is set up by a university to find out what students think about the quality of the education the school offers.

When using a group comparison to analyse the results, the researcher might first research the differences in scores between men and women. Next to “Select Split Variable:” he chooses “Gender”. Next, he’ll see that there are no significant differences for all the facets, between men and women; because there are only bullets in the main graphic. We must note that, however this conclusion isn’t wrong; the significance test isn’t performed between the results for men and women. Firstly, the test is performed between the mean result and the result for men; and secondly, the test is performed between the mean result and the result for women.

After this, the researcher might want to see if there is a significant difference to be concluded among the people who have different feelings in regard to the workload intensity. He would then select “Workload” for the split variable and see that there are quite a lot of significant differences to be found here:
  • There is a significant difference between the mean rating all the respondents gave the tests, and the rating the respondents who felt the workload was light gave the tests.
  • Again, there is a significant difference between the mean rating all the respondents gave the tests, and the rating the respondents who felt the workload was medium gave the tests.
  • Alternatively, there isn’t a significant difference between the mean rating all the respondents gave the tests, and the rating the respondents who felt the workload was high gave the tests.
Comparison Table

Group Comparison Facet graph and table
print
rating
  Comments
Copyright 2009 - 2012 by Kpiware - Powered by Webmove