Interpreting the 2006 Results

In reviewing the results presented, please keep in mind some of the following issues involving data analysis.

Variables Not Accounted for in the Analysis

The results presented are meant to describe the overall pattern of the data. They do not take into account or "control for" any characteristics of the respondents. For example, employee work satisfaction may be influenced by employee age. If we find differences between campuses on work satisfaction, if may be because there are differences in the age of workers on each campus. Although additional analyses can "control" for some of these effects, the current report presents the overall snapshot of results due to the large number of possible variables to control.

Statistically Significant Differences

Although the graphs may suggest differences among groups, in many cases these differences are not statistically – or meaningfully – different.

• For the job category results for staff, most of the differences are unlikely to be due to chance. However, these differences may be the result of other variables that have not been "controlled for."
• For the campus results for staff, most of the small differences between groups are likely to be due to chance fluctuations and the groups can be treated as equivalent to one another.
• For the campus results for faculty, the majority of the differences between campuses were statistically significant. That is, these differences are unlikely to be due to chance.
• In most cases, Crookston reported less favorable attitudes and experiences than their colleagues on the other campuses.

Practical Significance

Differences that may be statistically significant may not be practically significant. For example, one job category may have a score of 2.6 on work satisfaction and another may have a score of 2.4 on work satisfaction. This difference may be statistically significant, but not practically significant as they both show relatively favorable attitudes toward work. Practical significance rests on judgments of whether a difference is relevant or of any importance in the real world.

Sample Size and Variability

In some cases, the number of participants in a particular group is small. Results from a smaller group of participants are less stable and reliable as compared to results from a larger group of participants. Overall, standard errors (a simple measure of the degree of uncertainty of an average) were low (0.01-0.03) indicating that the different dimensions assessed in the Pulse Survey such as job satisfaction, pay satisfaction, etc. were sound estimates of the overall University population. Standard errors were slightly larger (0.08-0.20) for the Crookston and Morris campuses, due to their smaller sample size.