2x2x2 Design: Traffic x Cell phone use x Age DV: Brake onset time Light Traffic Heavy Traffic On Cell Phone Not on Cell Phone On Cell Phone Not on Cell Phone Younger 957 928 1112 933 Older 928 929 1412 1250 Does Traffic moderate cell phone use and age? - Main Effects: average across each level of the variable! (i.e. for age, add up ALL younger ones and divide by # of categories) - Two Way Interactions: o In Graphs: easy to tell. Compare the two lines that are grouped by the moderating variable, and see if they are parallel.
Final Exam Review - Three Way Interactions: if both two way interactions are different, then there IS a three way interaction! If there are no two-way interactions, there cannot be a three-way, and if the two-way interactions are the same, there cannot be a three-way! Threats to Internal Validity: - Two categories of things we can do to avoid threats: - By doing things right: o Design Confounds: situation in which some variable varies systematically with levels of your IV o Selection Effects: whether you’ve created equivalent groups in a between-group design correctly o Order Effects: (within-group designs) if you have not counterbalanced, you can get a finding that relates to order than your IV o Demand characteristics, experimenter-expectancy effects, placebo effects! By Using Comparison Groups: 1. Maturation: occurs when you’ve had a change on the DV and you are unsure whether the change would have occurred spontaneously or as a result of your IV. o I.e. for a treatment program for depression à get a pre-test score using BDI and then give therapy, then another BDI score. You see significant improvement. o However, you need a control group to see if it was in fact your IV that did it! If the control group also does better, we can see whether that is a result of a spontaneous event, and not due to our external event (our IV manipulation)! 2. History: over time, something external to the experiment has happened that has changed the participants. This may appear as being a result of the IV, but change is actually attributed to some historical event! o How do we know if history is responsible for the event? à Probably control group 3. Regression to the Mean: statistical concept where extreme scorers at time 1 are likely to be normalized at time 2. Any extreme performance will tend towards the average. o For example, a person is less likely to get 100 on a test the second time, as the first time may have just been the result of some random factors. o Because they’re random factors, it is unlikely that they will combine again in the same way! o Example A: if control group is significantly different, there is likely not regression to the mean o Example B: If the two groups started vastly apart, then come together we may say there is regression to the mean. o Crossover Effect: a comparison group can show us that there is regression to the mean, but then it crosses over, showing the treatment does have an effect! 4. Attrition: participants have been pre-tested, but then leave (for any reason) and you cannot post-test.
- Fall '15
- Dr. Ostovich