Class Notes (838,386)
Canada (510,872)
Todd Alway (64)
Lecture

October 9, 2013-3N06 Lecture Notes

6 Pages
81 Views
Unlock Document

Department
Political Science
Course
POLSCI 3N06
Professor
Todd Alway
Semester
Fall

Description
October 9, 2013 Political Science 3N06 O CTOBER 9, 2013 Political Science 3N06 2013 Lecture 5b: Experimental methods The effect of attack ads on voting E XPERIMENTAL G ROUP Pretest Independent Variable Post Test Dependent Variable Attack Ads Change in Dependent (Voting) Variable CONTROL GROUP Pretest Independent Variable Post Test Dependent Variable Must show them an ad Change in Dependent (Voting) (but not an attack ad)Variable • We can’t be very confident that the attack ads are responsible for the dependent variable • Does not have a lot of internal validity. • Subjects might be vulnerable to History/Maturation/ o History – War might make people o Maturation – People age, etc. variable that can affect the dependent variable. • What can we do to improve it? If we want to increase our internal validity? INTRODUCE A CONTROL GROUP. - Some of the limitations of a one group pretest/post-test design can be countered by using a control group o A group that should resemble the experimental group in every aspect but one – it is not exposed to the independent variable o The experimental and the control group must be equalized in every respect. o Equal in every respect, except for the fact that the independent variable will be applied. 1 October 9, 2013 Political Science 3N06 - Scores between the control group and the experimental group can be compared – with the difference being attributed to the independent variable - Using a control group can increase the internal validity of an experiment INTERNAL VALIDITY - Threats to the internal validity of an experiment continued: o How can you ensure that groups start out the same? 9. S ELECTION B IAS - Control groups only form a control when they are effectively equivalent to the experimental group - “Selection bias” (in this context) implies that the control and experimental group are not equivalent - If so, then the difference in scores in the post-test (between groups) may be reflecting the difference in subjects, not the effect of the independent variable that we are interested in W AYS TO ENHANCE PRE -TEST GROUP EQUIVALENCE A) Probability sampling - We will discuss probability sampling next lecture o To ensure pre-group equivalence, run the whole thing according to probability sampling (random sampling)  Random sample of Canada, one of the random sample groups go into the control and the other experimental  PROBLEM: It’s difficult to do in an experimental context. Getting people to volunteer is limited. Lab experiments rely on volunteers, and because of this there are certain limitations when it comes to the generalizing of the bias. B) Random assignment - Individuals in the sample are assigned to the control or the experimental group on the basis of random chance alone - As long as the groups are of a sufficient size, random assignment should equalize their average characteristics o Must use some random process in order to divide them. 2 October 9, 2013 Political Science 3N06 o Using chance and chance alone to divide between control groups and experimental groups is more likely to equalize of the group. o Must have at least 30 subjects; - SPSS (Percentage VS Valid Percentage) - In case of ordinal variables, only calculate the median. - Random assignment works better the larger the group is. C) Matching - Where groups are small, random assignment may not work to equalize the experimental and control groups - In that case, matching individuals into pairs (on the basis of shared characteristics) o And then splitting the pairs – with one member going into the control and one into the experimental group o May be the more appropriate option - Although even here – which characteristics should be relevant for the matching process? - Can you find subjects that are perfectly matched on all relevant variables? E XTERNAL VALIDITY - A second general consideration when evaluating what an experiment has actually demonstrated - The experiment may be internally valid o But how generalizable are your findings to the real world? o Is what is true in the laboratory true for society outside of the lab? o We want to know what causes things to happen in the real world; criteria for evaluating whether we’ve proven anything. Is it likely to tell us anything about the real world in which the lab took place? - Two questions that can guide your assessment of External validity 1. Can we generalize from our experimental participants to the general population that we are interested in understanding? 3 October 9, 2013 Political Science 3N06 a. Are the people that we conducted this experiment on similar to the rest of the population? b. What is it about the particular group that might create bias? 2. Can we generalize from the artificiality of the lab to the complexity of the real world (ecological validity) A) Mundane realism – is the
More Less

Related notes for POLSCI 3N06

Log In


OR

Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Sign up

Join to view


OR

By registering, I agree to the Terms and Privacy Policies
Already have an account?
Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.

Add your courses

Get notes from the top students in your class.


Submit