Ch 7

What are the major components of experimentally designed research
1. Independent variables and dependent variables
2. Pretest and Posttest
3. Experimental and control groups
What are the major components of Quazi experimental design research?
Experimental design research is used when researchers cannot meet the requirements of classical experimental design.
1. Non-equivalent group design: assignment of study participants not done through randomization. A matching process is used to attempt to make the experimental group (EG) and the control group (CG) (comparison group) as equal as possible. Example, Glueck and Glueck (1950) unraveling juvenile delinquency (500 delinquents and 500 match non-delinquents.
2. Time-series design: examine a series of observations on some variables over multiple points in time. An example is examining trends in arrests for drunk driving over time to see whether the number of arrests is increasing, decreasing, or staying constant.
What is the independent variable called in an experimental design
The independent variable in experimentation is called the “experimental stimulus”. It is administered, or given to some subjects (people or animals) but not others.
IVs in example: Cancer treatment, HGN, and mentoring.
DVs in example: tumor size, hitting performance, and future delinquency.
What do we call the measurement of the dependent variable before and after the introduction of the experimental stimulus
The dependent variable is measured prior to the administration of the experimental stimulus (pretest) and then measured again after the administration of the experimental stimulus (posttest).

Pretest example: Tumor size before experiment, batting average previous year, previous delinquency involvement.

Posttest example: Tumor size after experiment, batting average next year, delinquency next year.

What do we call the different groups that we assign people to in our experiments?
Those who are exposed to the experimental stimulus (IV) make up the “experimental group” of subjects, while those who are not exposed to the experimental stimulus make up the “control group”.

EGs in example: rats receiving cancer treatment, players receiving HGN injections, probationers receiving mentoring services.

CGs in example: rats receiving no treatment, players receiving no injection, probationers receiving typical services.

What are double-blind experiments
(see p. 189)
In medical experimentation, patients sometimes improve when they think they are receiving a new drug; thus it is often necessary to administer a placebo to a control group. In medical research, the experimenter may be more likely to “observe” improvements among patients who receive the experimental drug than among those receiving the placebo. A double-blind experiment eliminates this possibility because neither the subject not the experimenters know which the experimental group is, and which the control group is.

-Randomization: assigning participants to the EG and CG through a random process.
-Purpose: This helps to avoid any potential bias in how participants are assign to the groups and helps to ensure that those in the EG and the CG, are the group level, vary similar on various characteristics prior to the introduction of the IV
-Definition from the book: After recruiting a group of subjects under the classical experiment we would randomly assign those subjects to either the experimental or the control group. This might be accomplished by numbering all the subjects serially and selecting numbers by means of random-number table.
Random assignment is a central feature of the classical experiment. The most important characteristic of random assignment, sometimes referred to as randomization, is that is produces experimental and control groups that are statistically equivalent (same amount in both sides of the EG and CG). In other words, random assignment reduces sources of systematic bias in assigning subjects to groups. Random assignment to the EG and the CG helps rule out the possibility of selection bias.

What is randomization? What does it achieve?
-Randomization: assigning participants to the EG and CG through a random process.

-Purpose: This helps to avoid any potential bias in how participants are assign to the groups and helps to ensure that those in the EG and the CG, are the group level, vary similar on various characteristics prior to the introduction of the IV

-Definition from the book: After recruiting a group of subjects under the classical experiment we would randomly assign those subjects to either the experimental or the control group. This might be accomplished by numbering all the subjects serially and selecting numbers by means of random-number table.

Random assignment is a central feature of the classical experiment. The most important characteristic of random assignment, sometimes referred to as randomization, is that is produces experimental and control groups that are statistically equivalent (same amount in both sides of the EG and CG). In other words, random assignment reduces sources of systematic bias in assigning subjects to groups. Random assignment to the EG and the CG helps rule out the possibility of selection bias.

What are the major differences between experimental and quasi-experimental design?
The differences between experimental and quasi-experimental design is that experimental design is mostly achieve by randomization. Moreover, an experimental design may have more than one group receiving different versions or levels of experimental treatment. We can also vary the number of measurements made on dependent variables.

Four basic building blocks are present in experimental design: (1) the number of experimental and control groups,
(2) the number and variation of experimental stimuli,
(3) the number of pretest and posttest measurements, and
(4) the procedures used to select and assign them to the groups.

On the other hand, when random assignment is not possible, the next best choice is often a quasi experiment. In most cases, quasi experiment do not randomly assign subjects and, therefore, may suffer from the internal validity threats that are so well controlled in true experiments. According to the book, quasi experiment design is grouped into two categories: (
1) nonequivalent group designand
(2) time-series design.

What are the 4 different types of Time-series designs? What do they look like graphically?
a. Simple Interrupted: introduce an experimental stimulus and leave it in place.

b. Interrupted with non-equivalent comparison group: introduce an experimental stimulus and leave it in place, plus another location where there is no experimental stimulus introduced.

c. Interrupted with removed treatment: Introduce experimental stimulus at one point, and remove at later point.

d. Interrupted with switching replications: Introduce experimental stimulus at different points in time at different locations.

What is case-oriented research? Variable-oriented research?
Classical experiments and quasi-experiments with large numbers of subjects are examples of what Charles Ragin (2000) terms case oriented research, in which many cases are examined to understand a small number of variables.

Time-series design is examples of variable-oriented research, in which a large number of variables are studied for a small number of cases or subjects.

What is a case study
The case study design is an example of variable-oriented research. Here the researcher’s attention centers on an in-depth examination of one or a few cases on many dimensions. The term case and term study are used broadly. Cases can be individual people, neighborhoods, correctional facilities, courtrooms, or other aggregations.
What is a factorial design? A posttest only design? What do they look like graphically? (see figure 7.3 on p. 199)
A factorial design is a multiple experimental stimuli.

Posttest only design requires randomization.

-Because study participants are randomly assigned to the EG and the CG, we can assume that the two groups, on average, will be the same on characteristics like previous batting average.

What are the different types of threats to internal validity?
History, Maturation, Testing, Instrumentation, Selection bias, Experimental mortality
Give examples of each type of internal validity threat
1. History: Some external event may occur that is not part of the experimental stimulus that influences the posttest.
– Example: Imagine you are conducting some type of experiment in neighborhoods to reduce gun violence. A high profile shooting during the course of the experiment (e.g., Newtown shootings) may impact future gun violence, independent of the experimental stimulus. However, because experimental designs include a control group, people living in areas that are part of both the EG and CG should be influenced by the shooting event. As such, any additional change in gun violence in the EG is considered to be evidence of the effectiveness of the intervention.

2. Maturation: Everyone in an experiment will experience biological or psychological development, including aging, which may influence the posttest. Without a control group, it is hard to know if any behavioral change is due to the experimental stimulus or just getting older. Thus, the presence of a control group in experimental research accounts for and rules out any maturational effects.
Example: An intervention to increase self-control – people may naturally increase in self-control from adolescence to adulthood, so you need people in a control group to account for any maturational effect.

3. Testing: Study participants may realize the purpose of a study and report changes in the posttest that do not really reflect the effect of the experimental stimulus. This assumes that the pretest and posttest measures are based on survey data. Testing would be less of a concern for using officially recorded data, such as arrest records.
Example: A clinical researcher conducts an experiment by developing a new depression medication. Participants in the EG and CG self-report their feelings of depression at the pretest. Well, by the posttest, when they are asked to answer the same questions again, they might report less depressive symptoms because they think that’s what the researcher wants to hear. However, this should result equally for both the EG and CG participants. Without a CG, testing would be a serious issue.

4. Instrumentation: If the pretest and the posttest are measured differently, researchers may be detecting changes that did not result from the experimental stimulus, but rather the change in measurement. The bottom line – you shouldn’t change the way in which you operationalize the DV from the pretest to the posttest. We want reliable (consistent) measurements.

5. Selection Bias: How you choose who participates in a study can influence the results. Random assignment to the EG and CG helps to rule out the possibility of selection bias. Also be aware that recruiting volunteers for an experimental study may bias the results because volunteers are usually more motivated and more interested in a study than non-volunteers.
The issue of “creaming” or “cherry-picking” would be another example of selection bias. Imagine that a new public charter high school was opened, and rather than using a fair, random process to decide which children whose families applied got into the school, instead the school administrators picked those students who had the best elementary school grades. Well, of course the standardized test scores of the students attending the charter school would be higher than a regular public school. This is a great example of selection bias – the way in which students were selected results in bias.

6. Experimental Mortality: Same idea as panel attrition – individuals may drop out during the course of an experiment. This potential problem tends to increase as the time period between pretest, administration of the experimental stimulus, and posttest also increases. We cannot force individuals to continue to participate in an experiment. However, we can look to pretest information to see if those who drop out are different than those who remained in the study. Small amounts of compensation for participation and trying to convince people of the importance of the study can help reduce experimental mortality.

What are the different types of threats to External Validity?
1. Would the results of an experiment in a particular setting, at a particular point in time be found in other settings at other times? Studies done in well controlled environments (laboratory settings and simulated settings) may not apply to real world conditions. Think about the Paternoster et al. study you read for this week. A key issue is whether any given experiment fails to adequately simulate the conditions under which behavior is exhibited in the real world If so, then the external validity of the experiment may be quite limited. In other words, if the effects of an experiment only apply in laboratory setting, the relevance for the social world outside of the laboratory, and thus the external applicability of the experimental stimulus, is limited.
What are the different types of threats to Construct Validity?
1. One way to think about this issue is to think of dosage: Was there enough of the experimental stimulus to expect changes to take place. Was the drug given in a high enough dosage? Was the increase in police officers in the streets enough for criminals to notice? Was the rehabilitation program long enough to get individuals to reorient their thinking?

2. Just as we are concerned about the construct validity of the IV (the experimental stimulus), we are also concerned about how well we have measured the DV at both the pretest and posttest. If the DV is crime, what is the best way to measure crime? if the DV is self-control, did we create a valid measure of self- control? If the DV is trust in the police, were the questions used to measure trust adequate?

What are the different types of threats to Statistical Conclusion Validity?
1. The larger the number of individuals in an experiment, the greater the confidence we can place in study results. So having more individuals in an experiment reduces any threat to our statistical conclusion validity. Unfortunately, experiments can be quite costly.

2. The larger the change in the DV from the pretest to the posttest (think of this as the size of the correlation), the greater the confidence we can have that there is a true cause and effect occurring between the experimental stimulus and the DV. For example, let’s imagine that an intervention program is devised to reduce suicide. If 5% of people in the EG have committed suicide at a 10-year follow-up and 6% of people from the CG committed suicide, that difference is not much – there is little statistical conclusion validity of the effectiveness of the intervention. Now imagine that 0% of people from the EG committed suicide at the 10-year follow-up. That provides greater statistical conclusion validity.

Get access to
knowledge base

MOney Back
Guarantee
No Hidden
Charges
Unlimited
Knowledge base
Become a Member
Haven't found the Essay You Want? Get your custom essay sample For Only $13.90/page