October 28, 2010
- design an experiment
- today’s lecture helps prepare for assignment
- no collecting or analyzing data and not actually run an experiment
- explain, independent and dependent variable, stimulus, hypothesis and results to expect
- Rubric with TAs!
Experiments are the best way to isolate and make a causal argument. For surveys, there are always
other factors that may affect results
The Media Report
Ex: The Windsor Star. Breast feeding benefits mothers
Educational Implications of Stereotype Threat
Research by: Michael Inzlicht
Stereotype Threat (ST)
- ST is the discomfort individuals feel when they are at risk of confirming a negative stereotype of their
Ex. The only black guy in a group of whites
- This discomfort can results in poor intellectual experiences (ethnicity written on top or bottom of tests)
What is the relation between stereotype threat and being outnumbered?
Math test with groups of three.
Basic Features of the Classic Experiment:
1. Tests effect of independent variable (stimulus) on dependent variable
a. Ex: breast/bottle feeding is independent
b. Independent and dependent variable has to be operationally defined (how is it defined,
how are you going to measure it)
2. Pre-testing and Post-testing
a. Subjects are measured on their dependent variable before you give them the stimulus.
Then you measure the dependent variable again to see if it has changed
b. In breast-feeding example, there was no pre-test, only post-test
c. Maybe participants are more prepared or know what to expect and does better in post-
test than pre-test. So pre-testing may throw off the results in post-test; confusion
between stimulus or pre-tests is affecting
3. Experimental and Control Groups
a. Always have control groups that does not give stimulus
b. Hawthorne. If there had been a control group; one that is observed and has lights on
and one that is observed and has lights off. c. Always need to have control groups in medical research. Placebo effect.
Need to be careful of experimenters on affecting results. TA giving higher marks that she gave
comments to, etc.
- Researchers themselves don’t know who’s getting placebo or real drug because sometimes
experimenters are unconsciously affecting the participants if they knew.
1. Probability Sampling
a. Not done very often; not as necessary
b. Most probability sampling require 100 people, but in experiments, usually don’t need
that many people
2. Randomization (aka Random Assignment) – different from random sampling!
a. Randomly assign people into control and experimental group. (All characteristics should
* Can only be generalized from random sampling (Means we get our subjects however we can).
However, this is not as important in experiments. We’re looking at how people respond to something.
* To be talked about in paper, how we’re selecting our group and how we assign them.
Pre-Experimental and Quasi-Experimental Designs
1. One-shot Case Study
a. Generally have no pre-test and no control group
2. One-group Pretest Posttest Design
a. Aka as Before and After Design
b. Do a pre-test and a post-test, no control group. People serve as their own control group
because you’re comparing their pre and post test scores
3. Static-group Comparison
a. Do have an experimental and control group; no pre-tests.
b. No pre-tests so hard to know how much they differed prior to experiment.
4. Ex Post Facto Control Group Designs
a. Use control groups after the fact. Control groups not randomized; designated after the
b. Ex: Compare after the fact, study groups of different kind of compositions. (Michael
5. Nonequivalent Control Group Desig