Class Notes (1,100,000)
CA (630,000)
U of G (30,000)
SOAN (900)
SOAN 2120 (400)
Lecture

# SOAN 2120 Lecture Notes - Random Assignment, Selection Bias, Statistical Significance

Department
Sociology and Anthropology
Course Code
SOAN 2120
Professor
David Walters

This preview shows page 1. to view the full 5 pages of the document.
October 25, 2012
Experimental Designs
- independent variable dependent variable (differences in outcomes
between two groups)
- equivalent
o average scores on the dependent variable and
o scores on all other variables which are likely to affect the dependent
variable
- differences: manipulation
- how do we establish the equivalence of two or more groups?
o Random number tables
o Objective: equal probability of being assigned to the group
- Essential: (the groups are equivalent)
- Conclusion: manipulation
- Even with RA the two (or more) groups will not be exactly equal in all
respects.
- If the assignment is random, we can use tests of statistical significance
(manipulation vs. random variation)
- As n increases, individual differences decline
- Without random assignment (RA): findings may be attributable to pre-
existing differences between groups
o What is this called? Selection bias
- We could compare pre-test scores (dependent variable measures) before the
experiment begins
o Significantly different? “equivalent” proceed as if random
- What about all other potential confounding variables?
- This happens automatically with random assignment
Experimental Design Notation
- Standard Notation (page 181)
o R O1 X O2
o R O3 O4
- X represents an “experimental variable or effect” (the independent variable)
- O represents a dependent variable observation. Subscripts (scripts) depict
the observation; the exact ordering of the numbering is somewhat
unimportant.
- This experiment has a pre-test and post-test.
- First (the top group) gets the change.
- Second group is the controlled group.
- The pre-test for the experiment group (O1)
- The pre-test for the control group (O3)
- Finally, the letter R denotes RA

Unlock to view full version

Only page 1 are available for preview. Some parts have been intentionally blurred.

Example
Experimental Designs (Many Groups)
- Do “gimmick devices” improve drinking times?
- Group 1: “drink through a bottle” - 3.9 seconds
- Group 2: “regular bong” - 2.9 seconds
- Group 3: “bottle bong” - 2.5 seconds
- Different levels of the independent variable (different treatments)
- This experiment would be diagramed as follows:
- R X1 O1
- R X2 O2
- R O3
Notice that in this example, there is no pre-test
Does this invalidate any of the findings?
Internal Validity
Whether the experimental condition (treatment) actually had an effect within the
experiment itself.
- History
o When some event other then the experimental manipulation affects
an outcome, or changes respondents
o Before/after design
o How can we know that the change that occurred as a result of the
experimental manipulation, as opposed to all of the other things
which respondents experiences between the pre-test and the post-
test?
- Maturation
o Maturation is a subset of history
o With the passage of time, people get tired, get bored, get hungry
o Children subjected to a pre-test post-test design (before/after design)
Example:
Pretest: play with toys
Treatment: violent cartoon
Post-test: play with toys
o Changes due to cartoon? Or maturation factors as the afternoon wears
on?
- Testing
o Subjects are observed (tested) more than once
o Early testing influences later testing
o Example:
Does exposure to a cartoon will lead to more violent play-
behaviour?
Children are viewed through a one-way mirror playing with
toys (unobtrusive testing)

Unlock to view full version