1. What are the characteristics of true experimental designs? Preexperimental designs look good at first, but there are too many other explanations that can explain the results. These are bad designs and should be avoided. They cannot be interpreted (non-interpretable). Quasi-experimental designs are still not true experimental because the research cannot truly randomize. Also, it is too hard to control extraneous variables. True experimental designs are true because the subjects can be randomized, or they can use randomized matching.

2. What are the major true experimental designs? The major designs are: (1) randomized subjects, posttest-only control group design, (2) randomized matched subjects, posttest-only control group design, (3) randomized subjects, pretest-posttest control group design, (4) Solomon three-group design, (5) Solomon four-group design, and (6) factorial designs.

3. What are the strengths and weaknesses associated with each design? The randomized subjects, posttest-only control group design means that you have an experimental group and a control group. There is not a pretest, treatment is given to the experimental group, and the posttest is given to both groups. This design allows randomness with a random assignment of subjects to the two groups (chance favors that the two groups will be similar) and no pretest worries (no pretest sensitization, no worry about others figuring out what is going on, no maturation, and no history concerns because of the two groups). The treatment should have the same results, and the chief weakness is that the researcher cannot assess change.

The randomized matched subjects, posttest-only control group design is like the above design except that the researcher has matched randomized groups. The researcher would choose the above design over this design if he has a homogeneous population (no subgroups). If there is a special population, group, or characteristic that would affect the results, then the researcher should use this second design. However, unless there is a strong rationale for using the second design, the researcher should choose the first design.

The randomized subjects, pretest-posttest control group design is randomized and has a pretest. The strengths are: (1) can control maturation because of randomization, (2) history, and (3) can look at changed scores.

The Solomon three-group design is stronger than either of the above 3 designs. The researcher has 3 groups, can give some pretests and others not, and the results will show if the pretest has an effect or if the treatment does.

The Solomon four-group design, with four groups, is the strongest design. The Solomon three-group design is the first 3 lines of the below table. The Solomon four-group is all 4 lines. Lines 3 and 4 are the randomized subjects, posttest-only control group design. The Solomon four-group design is a combination of randomized subjects, posttest-only control group design (lines 3 and 4) and randomized subjects, pretest-posttest control group design (lines 1 and 2). The format of the Solomon four-group design is as follows:

The factorial design lets the researcher test for interactions. One treatment, for example, is more cognitive (treatment A), and the other (treatment B) is more about feeling. Not surprisingly, males would probably score higher on the first treatment and females on the second. The researcher can have males taking treatment A, males taking treatment B, females taking treatment A, and females taking treatment B. He would, then, do ANOVA to find out if the groups are different, if males did better than females, and it will tell him about the interaction for the counseling techniques and male or female. The weakness is with the different groups (it would not be the same males and females taking each of the two treatments, but randomization would be a good way to try to handle this).

4. What is causal-comparative research? This research is not really an experimental design. It tries to assess whether one variable is related to another "after the fact." This research lets the researcher infer whether or not an act or event in the past is affecting the variable. If one is studying women for self-esteem who have been victimized, then it is not possible to randomize the selection of women. The researcher must use the available subjects and then try to argue that being victimized affects self-esteem.

The steps to this research are: (1) establish a relationship between the independent (X) and dependent (Y) variables, (2) have two groups, (3) run self-esteem test, and (4) run a t-test to analyze the results. The researcher must show that X precedes Y in time and that no other factors can determine Y. Note that victimization occurs more and low self-esteem can be a function of socioeconomic status, so the researcher must try to factor that out.

5. What are possible problems associated with causal-comparative research? The possible problem is spurious results in the following forms: (1) common cause (is there a relationship between the number of drunks in a town and the number of churches? Spurious results might say that more churches make more drunks when really the common cause might be population.), (2) reverse causality (books read before kindergarten affect IQ. Then, in high school, testing might show that this is true. But the truth (reverse causality) might be that higher IQ'ed people read more books.), and (3) other possible variables that can explain the relationship. The researcher must rule out other possibilities.

6. How can we approximate control in causal-comparative research? The researcher can (1) look at changed scores to establish differences between the two groups (implement a counseling technique and measure self-esteem again (this moves into a quasi-design technique)), (2) can involve matching (put different levels of socioeconomic people into both groups for the women/self-esteem test), (3) use ANCOVA to parcel out any influence or variable effect (Analysis of Covariance), (4) use homogeneous groups, and (5) build extraneous variables into the study (try to have the same number of low, middle, and high class women in each of the two groups).

7. What are the strengths and weaknesses of causal-comparative research? The strengths are: (1) useful when true and quasi-experimental design will not work, (2) cannot manipulate attribute variables (male and female are assigned), and (3) recent information and improvements have strengthened this design (create a path analysis). The weaknesses are: (1) cannot manipulate the independent variable, (2) cannot control nuisance variables, and (3) cannot establish cause and effect because the researcher cannot carry out the experiment. He or she can say that the evidence supports, but they cannot say anything concrete for sure.

				Tom of Bethany

Back To TLEE's Home Page

Index to Selected Essays And Book Reviews

Lesson 24. Descriptive and Survey Research

Reaction to Video Lesson 21

Send email to: tlee6040@aol.com