Experimental Variable

-manipulated or varied by researcher

- is ALWAYS an independent variable

R

random assignment

G

Group

X

Experimental Variable (independent)

O

Outcome variable (dependent)

experiemental variables

levels of experimental treatment

General Criteria for a well Designed Experiment

Essentially the characteristics that make for a good research design also apply to the design of an experiment.

1. Adequate experimental control

2. Lack of artificiality (i.e., meaningful application)

3. Has a comparison group

4. Adequate data to allow for statistical precision

5. Uncontaminated data

6. No confounding variables

7. Representativeness

8. Parsimony – mindful of number of variables explored

Construct Validity

Applies to the operational definitions of variables in experiment (construct); concerns that they could be construed for some other construct.

The extent to which a test measures one or more dimensions of a theory or trait.

Statistic Conclusion Validity

Applies to the statistical analysis and whether or not there exists a statistically significant difference between the experiment and control group.

Threats to Internal Validity

1. History - fire drill so missed class

2. Maturation -

3.Testing (memory) - remembers questions from last test

4.Instrumentation - same post test given, one doesnt follow directions

5. Statistical Regression

6.Selection

7.Mortality* - dropping out

8. Selection-maturation interaction- JHS Students fatigue, HS Student do not during test

Threats to External Validity

**ON EXAM**

1. Interaction effect on testing - pretest effects

2. Interaction effects of selection biases and the experimental treatment - an effect of some selection factor of intact groups interacting with the experimental treatment that would not be the case if the groups were formed randomly

3.Reactive effects of experimental arrangements- effect due to artificial test setting

4.Multiple-treatment interference - potential carryover effect of treatments, results cannot be generalized to single treatment.

Threats to Construct Validity

1. Inadequate preoperational explication of constructs*

2. Mono-operational bias

3. Mono-method bias

4. Hypothesis-guessing within experimental conditions*

5. Confounding constructs and levels of constructs

* All within researchers control

Threats to Statistical Conclusion Validity

1. Low statistical power - using sample size that is too small to detect differences

2. Violated assumptions of statistical test

3. Fishing and the error rate problem-capitalizing on chance findings

4. Reliability of measures - using technically inadequate measures

Post-test only control group

contains as many groups as there are experimental treatments, plus a control group

Subjects are only measured after the experimental treatments have been applied

R G1 X1 O1

R G2 X2 O2

R G3 X4 O3

R G4 X4 O4

R G5 -- O5

Pretest - posttest control group design

contains as many groups as there are experimental treatments. plus a control group

Subject are measured before and after recieving the experiemental treatments

R G1 O1 X1 O2

R G2 O3 X2 O4

R G3 O5 X3 O6

R G4 O7 -- O8

Solomon Four-Group Design

tests on the effects of pretesting

is a combination of the posttest-only control group design and the pretest-posttest control group design

four groups included only on experiment used

R G1 O1 X O2

R G2 03 -- O4

R G3 -- X O5

R G4 -- -- O6

Factorial Design

involves two or more independent variables

R G1 X1 Y1 O1

R G2 X1 Y2 O2

R G2 X2 Y1 O3

R G3 X2 Y2 04

Repeated Measures

the same subject is measured more than once on the dependent variable

R G1 X O1-O2-O3

R G2 -- O4-O5-O6