Reading...
Play button
Play button
Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
23 Cards in this Set
- Front
- Back
|
Types of validity
|
• Internal
• External • Construct • Statistical Conclusion |
|
|
Internal Validity
|
• To what extent can the intervention (or condition) alone be considered to account for the results, changes, or group differences? The more it can be done, the greater the internal validity.
OR • Are the results (DV) due to variations in the IV? The better the experiment, the more it rules out possible alternative explanations of the results. |
|
|
Factors other than the IV which could be responsible for the results are called:
|
Threats to Internal Validity
|
|
|
Threats to Internal Validity:
1. History |
• Any event inside (except the IV) or outside of the experiment that may account for the results.
• Refers to events shared by all or most subjects in all groups. Ex. an earthquake could increase all or more subject's anxiety |
|
|
Threats to Internal Validity:
2. Maturation |
• Refers to changes that result from processes internal within the participants. e.g. getting older
• Is a problem only when the effects of maturation cannot be separated from the effects of the intervention. |
|
|
Threats to Internal Validity:
3. Testing |
• Effects of practice on the dependent measures
• A group that receives repeated testing (e.g. pre and post) without receiving the intervention can help rule out this threat. |
|
|
Threats to Internal Validity:
4. Instrumentation |
Instrumentation
• Changes how the DV is measured over time. • Ex. scale may not be consistently accurate = not same pre/post • Threats to int. val. can occur when any of the following is not constant: - Measuring instruments - Observers, raters, or interviewers - Remarks or directions from the experimenter |
|
|
Threats to Internal Validity:
5. Statistical Regression |
Statistical Regression
• Extreme scores tend to change towards the mean over time • Ex. Very anxious people might be expected to calm down over time • No treatment control group allows the researcher to get an idea of how much the anxiety would decrease without treatment. |
|
|
Threats to Internal Validity:
6. Selection Bias |
Selection Bias
• Difference between groups because of selection or because of assignment to groups or treatments • One form is: the use of different methods for selecting participants for different conditions • Ex. studying the effectiveness of treatment after an earthquake. Treatment group seeks treatment; control group consists of people contacted by their church. • When there is selection bias, the groups were different before any experimental manipulation or intervention. • Random assignment of individuals to groups is usually used to avoid selection bias. • Point is you don't want the groups to differ before the intervention. • Often a problem when use groups that already exist (e.g., wards, classes). |
|
|
Threats to Internal Validity:
7. Attrition |
Attrition
• Participants dropping out of a study • Especially a problem if different types of participants (older, younger, healthier) drop out of different groups • Problematic in studies that last a long time; in those where intervention is unpleasant; in studies where one treatment is more pleasant than another (fewer people drop out of the good one) |
|
|
Threats to Internal Validity:
8. Combination of Selection and other Threats |
Combination of Selection and other Threats
• When a threat to internal validity differs for different groups - one group has an experience that the other group(s) did not have • Ex. if comparing therapy and control groups, can't conclude therapy was responsible for improvement in therapy group if participants in the control group experience more stress during the six months of therapy than participants in the therapy group. • Selection most commonly interacts with history or maturation |
|
|
Threats to Internal Validity:
9. Diffusion or Imitation of Treatment |
Diffusion or Imitation of Treatment
• When there is a tx group and a control group, some or all of the participants in the control group may inadvertently receive some or all of the tx, with the result that the difference between groups is smaller than it would be with no diffusion. • Ex. Half of the children at a school receive a brief course in avoiding sexual abuse. they were taught what kind of bx to look for in adults and how to respond to such bx. The students in the control group did not receive the course, but the children in the tx group tell the kids in the control group what they learned. |
|
|
Threats to Internal Validity:
10. Special Treatment or Reactions of Controls |
Special Treatment or Reactions of Controls
A. When the control group is given something so that they wont feel snubbed, that something may have an effect (e.g., money) B. 1. Participants in tx group may perform better soley because they know they are in a tx group so they try harder B. 2. Participants in a control group may perform better to show they are as good as the participants in the tx group B. 3. Participants in a control group may perform worse because they feel let down (because they are in the control group) If any of the above occurs: A. Differences will be smaller than if control group got nothing B. 2. Differences will be smaller than if control group did not change their bx B. 3. Differences between groups may be exaggerated |
|
|
General notes:
|
• Most threats to internal validity can be seen in case studies and in studies where there is only one group (a tx group). The changes over time cannot be attributed to the intervention.
• Threats to internal validity are often the alternative explanations of the results |
|
|
If a sample is is to provide useful descriptions of the total population, it must contain:
|
• The same variation as exists in the population
• Best accomplished through probability/random sampling • All members of the population have an equal chance of being selected |
|
|
Systematic Sampling
|
• Generally as good as random sampling
• Take every K-th person from your list. (A problem can occur if there is a cyclical pattern in the list that coincides with the sampling interval.) |
|
|
Stratified Sampling
|
• Use Stratified Sampling when we want to be sure that a sample is representative on certain variables, like gender or ethnicity.
• Members of groups in the population are selected in proportion to their representation in the population. |
|
|
Oversample
|
• Sometimes researchers oversample so that they have enough participants in each of a number of specified groups so that comparisons can be made among the groups.
• Ex. Oversample Native Americans so that comparisons can be made between them and other ethnic groups |
|
|
Accidental (Available) Samples
|
Take cases that are available until reach a specified N.
• Ex. The first 100 people on the street who are willing to be interviewed |
|
|
Quota Samples
|
Accidental samples where numbers of males and females, for example, are interviewed.
• Try to interview types of people in proportion to the presentation in the pop. |
|
|
Purposive Samples
|
• Pick cases that are judged to be typical of the pop of interest
• Used to forecast elections • Ex. For each state, select a number of small election districts whose election returns in previous years have approximated overall state returns - don't know if these will still be predictive. |
|
|
Cluster Sampling
|
• Because it is difficult and expensive to get random or stratified random samples, large-scale surveys often use this method.
• Procedure is: first sample groupings or clusters; then sample individuals from these • Can sample this way randomly |
|
|
Cluster Sampling Example
|
To sample seventh-grade public school students in CA
1. Select a random sample from a list of school districts 2. For the school districts chosen, then select a random sample of schools 3. Take a random sample of the classes in those schools (or all classes) 4. Survey all or a random sample of the children in these classes • Because there are stages in this type of sampling, it is called multistage sampling |
