2/10/2006

• Q: Genie Moore: use of historical control in animal study for secondary hypothesis testing
• A: We need to be careful to use historical control and avoid a possible bias due to the difference between populations. Considering that: (1) the study is a well controlled animal study; (2) the only condition which are different could be tested using another control group (pregnant saline treated dogs) wheter this condition affects the outcome of interest; and (3) the historical control will be used for thes econdary hypothesis testing, it would be O.K. to use historical control as long as they precisely specify and are awared of the possible bias which might affect the conclusion to make.
• Q: Dniel Moore: do animal experiment within one day vs. spread over time
• A: Unless they expect a day factor affects the outcome of interest or are interested in estimating day-to-day variability, it is better to do the experiment within one day since it would reduce variablity and more powerful.
• Q: Charlie Cox: how to test whether several measurements would differ by gender and by left and right
• A: Since the measurements are repeatedly measured on each subject, use repeated ANOVA or a regression model taking into accout correlation