Work perception and satisfaction ratings were not included, and overall ratings (when provided) were excluded since the focus of the MTMR matrix is on traits (dimensions).
This was not desirable in the present study since MTMR performance appraisal studies tend to have values below .4, and the focus was on evaluating the three models for typical MTMR studies.
One study (Lance et al., 1992) provided the MTMR matrix for first-order factors.
To address issues regarding the design of MTMR studies the following variables were recorded for each matrix.
These results indicate that with MTMR performance appraisal data, it is somewhat likely that method variance will be multidimensional.
The evidence strongly indicates that the CU model is superior to the other models for MTMR performance appraisal data.
The first is that the CU model is the most appropriate of the three models for MTMR performance appraisal data.
The superiority of the CU model and poor performance of the DP model is contrary to Goffin and Jackson's (1992) suggestion that the DP model was most appropriate for MTMR data.
In MTMR studies I believe the best way to ensure low method correlations is to reduce the proportion of method variance in ratings.
A third recommendation is that researchers strive to maximize the proportion of trait variance in MTMR data (minimizing the proportion of method variance is also recommended, and has been discussed above).
Findings regarding proportions of trait and method variance are potentially useful whenever performance appraisal systems are developed, whether or not MTMR research is conducted.