Refer to the Human Factors (Apr. 1990) study of recognizer accuracy at three levels (90%, 95%, and 99%) and vocabulary size at three levels (75%, 87.5%, and 100%) on the performance of a computerized speech recognizer, Exercise 12.26 (p. 661). The data on task completion times (minutes) were subjected to an analysis of variance for a 3 × 3 factorial design. The F test for accuracy × vocabulary interaction resulted in a p-value less than .0003. 

a. Interpret the result of the test for interaction.

b. As a follow-up to the test for interaction, the mean task completion times for the three levels of accuracy were compared under each level of vocabulary. Do you agree with this method of analysis? Explain. 

c. Refer to part b. Turkey’s multiple comparison method was used to compare the three accuracy means within each level of vocabulary at an experiment wise error rate of α = .05. The results are summarized here. Interpret these results.

Transcribed Image Text:

Accuracy Level Vocabulary Size 99% 95% 90% 75% 15.49 19.29 22.19 87.5% 12.77 14.31 16.48 100% 8.67 9.68 11.88