Lecture 16: Factorial Crossing Chapter 6 Frog Dissection Experiment Factorial Crossing. We have actually seen this in the experiment comparing different methods of frog dissection. Experimental Units: Lab sections of biology. Response: Average score on a test on the anatomy of the frog. 1 Frog Dissection Experiment 2 Factors and Levels Factor 1: Real Dissection Treatments: Real, Virtual, Both, None. These treatments are really the combination of two factors of interest. Levels: Yes, No Factor 2: Virtual Dissection Levels: Yes, No 3 Treatments 4 Treatments With factorial crossing treatments consist of all possible combinations of the levels of the factors. Real Dissection? Yes Real Dissection? No 5 Virtual Dissection? Yes Virtual Dissection? No Both Real Virtual None 6 1 Lecture 16: Factorial Crossing Why use factorial crossing? Doing one experiment using factorial crossing is more efficient than doing separate experiments, one for each factor of interest. Why use factorial crossing? Doing one experiment using factorial crossing allows for the investigation of possible interaction between the factors. 7 Efficiency 8 Two Experiments Experiment I: Compare Real Dissection to No Real Dissection. Experiment II: Compare Virtual Dissection to No Virtual Dissection. Alpha = 0.05 Beta = 0.10 1.4 standard deviation in means. 9 Sample Size 10 One Experiment Experiment I: 2 Groups, 12 per group, 24 lab sections. Experiment II: 2 Groups, 12 per group, 24 lab sections. Total of 48 lab sections. Alpha = 0.05 Beta = 0.10 1.4 standard deviation in means. 11 12 2 Lecture 16: Factorial Crossing One Experiment Different Means Experiment: 4 Groups, 16 per group, 64 lab sections. How is this more efficient? There is a problem as the two situations are referring to different means. Two experiments: difference in factor level means. One experiment: difference in treatment means. 13 Sample Size 14 Real Dissection Virtual Virtual Dissection? Dissection? Yes No Real Dissection? Yes Both n=6 Real n=6 12 Real Dissection? No Virtual n=6 None n=6 12 12 12 Comparing real dissection to no real dissection there are 12 lab sections per group. Can detect a 1.4 standard deviation difference in factor level means, Alpha = 0.05, Beta = 0.10. 15 Virtual Dissection 16 Greater Efficiency Comparing virtual dissection to no virtual dissection there are 12 sections per group. Can detect a 1.4 standard deviation difference in factor level means, Alpha = 0.05, Beta = 0.10. One experiment using 24 lab sections can detect the same size difference in factor level means as 48 lab sections used in two experiments. 17 18 3 Lecture 16: Factorial Crossing Important Distinction Important Distinction When there was only one manipulated factor, the factor levels were the treatments. Now that there are several factors there is a difference between factor levels and treatments. 19 20 Sample Size Interaction When using the sample size tables be sure to note the distinction between factor level means and treatment means. In Stat 301 or 401, two explanatory variables are said to interact if the relationship between the response variable and one of the explanatory variables changes for different values of the second explanatory variable. 21 22 Interaction Interaction In Stat 402, two factors are said to interact if the effect of one factor on the response is different for different levels of the second factor. The two definitions are getting at the same idea just with different terminology. 23 24 4 Lecture 16: Factorial Crossing Example Yes:Virtual Dissection In the experiment with Real Dissection; Yes, No and Virtual Dissection; Yes, No as the factors what would the treatment means look like with and without interaction? Average Score on test of Frog Anatomy No: Virtual Dissection No Yes Real Dissection 25 26 No Interaction In sections with no virtual dissection the effect of real dissection is the same as in sections with virtual dissection. Average Score on test of Frog Anatomy Yes:Virtual Dissection No: Virtual Dissection No Yes Real Dissection 27 Interaction 28 Interpretation In sections with no virtual dissection there is a positive effect due to real dissection. In sections with virtual dissection there is no effect due to real dissection. The effect of real dissection on the average score on the frog anatomy test is different for sections using, or not using, virtual dissection. Real dissection and Virtual dissection factors interact. 29 30 5