1. What are the names and definitions of the threats to internal validity? The threats to internal validity are: (1) history (an event that occurs during the experiment that affects the results, thus creating a rival hypothesis), (2) maturation (growth occurs in on of the groups being studied (best way to protect against this is randomization)), (3) testing (people become tired of the same test, so they give less and less effort), (4) instrumentation (changing instruments between pretest and posttest (raters can also change between these tests)), (5) statistical regression (novice researchers often choose groups based on extreme scores. Then, when the researcher tests, the incorrect conclusion is that the groups are alike because of the test. But this is not correct because of statistical regression towards the mean. The further one is from the mean, the more likely they will get closer with a retest.), (6) selection (treat and retest. The control group does not do well but the exoerimental group does (the one getting the treatment). Without random selection, the two groups were not alike at the beginning.), (7) mortality (loss of subjects over time (people die, lose interest, drop out, move away), (8) ambiguity of causal relationship (correlational studies – two characteristics vary, more of one than the other, but the researcher does not know if they have affected each other), (9) diffusion and imitation of treatments (a person in the control group finds out what is being tested and tells others, thus affecting the results), (10) compensatory equalization of treatments (having a traitor helping collect data but gives information to help the control group so that differences between the two groups cannot be detected), and (11) compensatory rivalry (the control group finds out that it is the control group, so they work harder to not be the dummies. If both groups do well, the researcher would not know why). These types of threats apply more toward the experiment itself, where the experiment is protected by a strong internal design.

2. What are the three types of external validity? The three types of external validity are: (1) population validity (will the results generalize to the population of interest, can chosen sample represent what the population will do, and need random sample with large sample size), (2) ecological validity (will the results generalize to other settings), and (3) external validity of operations (will the results generalize to different definitions and/or procedures, are they generalizable to other people).

3. What are the names and definitions of the threats to external validity? The threats to external validity are: (1) reactivity of outcome assessment (where subjects react to what they think is being studied (hypothesis guessing messes up the instrument)), (2) pretest sensitization (pretest tunes subject into what is being tested), (3) posttest sensitization (subjects figure out what the researcher is doing in the posttest causing different answers to the questions), (4) generality across construct and measures of outcome (treatment seems to work because of the instrument used to measure the effect, but another valid instrument would show a different outcome), (5) reactivity of experimental arrangements (subjects change just because they are being studied (Hawthorne effect) and not because of the treatment), (6) multiple treatment interference (treatment applied in wrong dosages can affect the results (a residual drug compounds itself and skews the results), (7) novelty effects (subjects react because of new stimuli (an example is the red brake light in the back window of cars where drivers react to the "new" light and not just to the light)), (8) experimenter effect (the experimenter can affect the results), (9) different operational definitions of the variable (different results are based on the definitions of the variables), (10) combination of selection and treatment (different groups get a treatment that matches them well causing favorable results because of the groups and not because of the treatments), (11) combination of experimental setting/situation and treatment (affects different groups in ways that give misleading results), (12) combination of history and treatment (the occurrence of a historcial event intercombined with the treatment can cause erroneous results), and (13) time of measurement and treatment effects (treatment may be more effective because of the time of day.).

				Tom of Bethany

Back To TLEE's Home Page

Index to Selected Essays And Book Reviews

Lesson 20. Experimental Designs (Part I)

Reaction to Video Lesson 19

Send email to: tlee6040@aol.com