Using previous work on the Behrens-Fisher problem, two approximate degrees of freedom tests, that can be used when one treatment is individually administered and one is administered to groups, were developed. Type I error rates are presented for these tests, an additional approximate degrees of freedom test developed by Myers, Dicecco, and Lorch (1981), and a mixed model test. The results indicate that the test that best controls the Type I error rate depends on the number of groups in the group-administered treatment. The mixed model test should be avoided.
Wehry, Stephanie and Algina, James
"Type I Error Rates Of Four Methods For Analyzing Data Collected In A Groups vs Individuals Design,"
Journal of Modern Applied Statistical Methods: Vol. 2
, Article 13.
Available at: http://digitalcommons.wayne.edu/jmasm/vol2/iss2/13