5.2 Principles of Experimental Cell Phone Use and Driving  We want to see if talking on a hands-free cell phone distracts drivers. We have a simulator and 40 randomly chosen undergraduate students. We need to design an . We will use braking time to determine how distracted they are. Randomized Comparative  This is the type of experiments we will be dealing with in this class.

produces 2 groups of subjects that we expect to be similar in all respects before the treatments are applied.

 Comparative design helps ensure that influences by variables, other than the one we want to measure, are equal on both groups.

 This lets us assume that the difference in our variable of interest is due to our treatment. Goal of experiments: To Find  This that the observed effect is so large that it is very unlikely to occur by random chance.

 Tells us that the treatment we are testing is having an actual effect on the subjects. Experimental Design Basic Principles of statistical design of experiments:  Control the effects of lurking variables on the response (compare 2 or more treatments)  Replicate each treatment on many units to reduce chance of variation in the results  Randomize—use impersonal chance to assign experimental units to treatments Experimental Design  Completely randomized design—all units are assigned at random.  Each unit has the same probability of being chosen.

 Our cell phone experiment is an example of this. Experimental Design  —group of units that are known to be similar before the treatments.

 This would be used if we know that the drug we are testing effects men and women differently.

 Form blocks based on most important, unavoidable sources of variability.

 We use this to mitigate variables that we cannot control and do not want to measure.

 This is only used when the variable has an obvious effect on the response variable. It will not always be gender. Soy-Bean Yields

 We have 2 farms that we can use for testing the effect of tillage type (two types) and pesticide application (three types) on soy-bean yields.  We need to use both farms in order to have enough samples.  We know that soil type and fertility vary greatly by location. Experimental Design  Matched Pairs design —block 2 units together (one control; one treatment) based on variables (that you’re not interested in testing) that you believe will effect the response variable.

 Match units with another unit similar to it.

 You can be matched with yourself.

 Note: The two units are not independent of each other, since they were paired specifically. Cell Phone Experiment - Revisited  The driving-cell phone experiment was actually done with a matched pairs design. Each subject was paired with themselves.  Double blind experiment —neither subjects nor those who measure the response variable know which treatment a subject received  Helps treat every subject the same, regardless of treatment

 Lack of realism —cannot realistically duplicate conditions we want to study Keep in Mind  Statistical analysis of an experiment cannot tell us how far the results will generalize to other settings.

 This is why being sure to select our sample appropriately, from our population of interest is very important. Example: Ultra-marathon runners tend to develop respiratory infections after the race. Will 600 mg of vitamin C daily reduce these infections? Researchers randomly assigned runners to receive either vitamin C or a placebo. Separately, they also randomly assigned these treatments to a group of non-runners. McDonalds Example

Do consumers prefer the taste of a cheeseburger from McDonald’s or from Wendy’s in a blind test in which neither burger is identified? Describe the design of a matched pairs experiment.