N503: Week 4 Sept. 28, 2005

1
N503 Week 4Sept. 28, 2005

• Experimental and
• Quasi-experimental designs

2
Essential components of true experimental designs

• Random assignment to groups
• Control of extraneous variables
• Manipulation of the independent variable
• At least two groups.

3
Basic design

• Example
• What is the effect of preop teaching on postop
  pain and medication consumption?
• Independent variable Preop teaching
• Dependent variables postop pain and medication
  consumption

4
Variations on basic design

• More than 1 variation in independent variable.
• Usual preop teaching, structured preop teaching
  with lecture, structured preop teaching with
  video
• More dependent variables
• Patient knowledge, postop pain, medication use

5
After only design

• R X O1
• R C O2

6
Example 1 Among adults, does an apple a day
keep the doctor away?

• 1. Randomly select a group of adults
  (generalizability)
• 2. Randomly assign a group of people to
  experimental or control groups (control)
• 3. Give people in the experimental group an
  apple each day for 6 months
• 4. Count the number of ill episodes for which a
  doctor visit was necessary in each group
• What is the independent variable?
• What is the dependent variable?

7
Before-After Design

• R O1 X O2
• R O3 C O4

8
Example 2 Do teenagers who no longer have
access to the family vehicle have more tidy
bedrooms?

• 1. Randomly select a group of teenagers
  (generalizability)
• 2. Randomly assign them to experimental or
  control (control)
• 3. Assess each student on a tidiness scale by
  rating their bedrooms
• 4. Give the experimental group an intervention
  that consists of removing access to the family
  vehicle
• 5. Assess tidiness a second time in both groups
• 6. Check whether the students who lost access to
  the family car had improved tidiness scores,
  compared to students in the control group.
• Independent variable?
• Dependent variable?

9
Solomon Four Group Design

• R O1 X O2
• R O3 C O4
• R X O5
• R C O6

10
Example 3 Does the provision of donuts at report
increase alertness in staff nurses?

• Randomly select group of staff nurses
  (generalizability)
• Randomly assign to one of four groups (control)
• Do assessment of alertness in 1 experimental and
  1 control group.
• Provide both experimental groups with donuts
  prior to the start of report.
• Now assess all 4 groups for alertness.
• Compare
• Scores of the after only with the scores of
  pre-test groups
• Scores of two pretested to each other
• Scores of two after-only groups to each other
• Scores of two experimental groups together to two
  control groups
• Indpendent variable?
• Dependent variable?

11
Essential components of quasi-experimental designs

• Manipulation of the experimental variable
• No random assignment to groups, and/or
• Extraneous variables added as independent
  variables
• Homogenous samples
• Matching
• No control groups
• Non-equivalent group designs (comparison group)
• Time series designs (own control)

12
Example 4 Does an apple a day keep the doctor
away? (time series)

• 1. Randomly selected participants
• 2. Randomly assigned to groups
• 3. Apples given to experimental group but some
  people didnt like apples and they gave them to
  people in the control group.
• 4. Retrospectively collected dr. visit data each
  month for 4 months prior to provision of apples
  and for 4 months following provison of apples.

13
Example 5 Does an apple a day keep the doctor
away? (non-equivalent group)

• 1. Random selection
• 2. Participants from 2 communities
• A. Community 1 is in an apple growing district
• B. Community 2 is in a district with no access
  to apples
• 3. Assess doctor visits at time point 1
• 4. Assess doctor visits at time point 2

14
Controlling unwanted influences in experimental
and quasi-experimental designs (internal validity)

• Extraneous variables
• Intervening
• Confounding
• Bias
• Hawthorne Effect
• Time

15
Data analysis for quasi-experimental designs

• Non-equivalent group designs
• ANOVA
• ANCOVA
• Time series designs
• Time series analysis (similar to regression)