Exam Flashcards
1
Q
Observational research
A
- In much of clinical research, subject characteristics and other variables are not manipulated directly by the investigator
- The variables are “manipulated by nature” and the investigator evaluates the impact of these variables through selecting persons for study who have the characteristic of interest
- Such studies are referred to as “observational research” to convey that the investigator is to observes (assesses) different characteristics and their associations, rather than to intervene experimentally
2
Q
Why observational research?
A
- In psychology observational research plays a special role because many core questions do not permit ethical experimental manipulation
o We are interested in individuals with special experiences due to exposure (e.g., to trauma, war, domestic violence, prenatal cigarette smoking) or to deprivation (early loss of parents, malnutrition)
o Obviously, one cannot assign individuals to experience one condition versus another. However, individuals with these varying characteristics can be identified and studied
o Experimental research, is often restricted to the manipulation of one or two variables at a time
o Multiple variables might influence a phenomenon and these variables may be related in dynamic (constantly changing), interactive, and reciprocal ways
o Observational studies can take into account multiple variables, study them over time, and examine the influences of variables on each other
o Data-analytic techniques have advanced over the past decades that can strengthen the inferences drawn from observational research
3
Q
Case-Control Designs
A
- Investigation of a characteristic of interest by forming groups who vary on that characteristic and studying other current or past features of those groups
- The key characteristic is in identifying groups who vary in the outcome (criterion) of interest
- Case-control design is the term used where “case” typically means someone who has the disease or condition (e.g., heart disease, high blood pressure) that is to be studied
- The investigator compares subjects who show the characteristic (cases0 with individuals who do not (controls). The independent variable is the characteristic that served as the basis for selection
- The investigator compares the two groups on the measures of interest and then interprets the differences to reflect a critical facet of the problem
4
Q
Cross-Sectional Case-Control Design
A
- In a cross-sectional, case-control design, the most commonly used version, subjects are selected and assessed in relation to current characteristics
- The goal is to examine factors that are associated with a characteristic of interest at a current point in time
- The investigator typically begins with hypotheses about how groups differ
- Subjects are identified and assessed on multiple characteristics beyond those used to delineate their status as cases or controls
- Because all of the measures are obtained at the same point in time, the results are correlational, i.e., one cannot know from the study whether the outcome preceded or was caused by a particular characteristic
5
Q
Retrospective, Case-Control Design
A
- The goal is to draw inferences about some antecedent condition that has resulted in or is associated with the outcome
6
Q
- As an example, a retrospective case-control design was used to evaluate the relationship of breastfeeding and ADHD in children
A
o Selective recall, inaccurate recall, and recall biased by the outcome (e.g., dysfunction) all interfere with drawing valid conclusions about the past event, its occurrence, or differential occurrence for groups that vary in a later outcome
o In some cases, historical records (e.g., school truancy, participation in high school activities) are used as the data. With such records, the quality, reliability, and the completeness of the data also raise potential interpretive problems
7
Q
- Not all retrospective reports are subject to the same sorts of degrees of bias
A
o Retrospective reports of psychological states (e.g., family conflict, mental health, difficulties of childhood) and duration, level, and dates of particular events are rather poor
o Recall of discrete events (e.g., changes in residences) and more stable characteristics (e.g., reading skills) tends to be more reliable but still not great – e.g., polysubstance interview
8
Q
Strengths of Case-Control Designs
A
- Well suited for studying characteristics that are relatively infrequent in the population
- Efficient in terms of resources and time because of the cross-sectional assessment
- No attrition because of assessment at one point in time
- Can study magnitude type of relations among variables (e.g., direct influence, moderating influence)
- Allows the investigator to match subjects on one of the variables assessed at pretest that may influence the results
- Can rule out or make implausible the role of influences that might be confounded with the characteristic of interest
9
Q
Weaknesses of Case-Control Designs
A
- No time-line is shown among the variables of interest, so one cannot usually establish whether one characteristic preceded the other or emerged together
- Causal relations cannot be directly demonstrated, even though various analyses (e.g., dose-response relations) can provide a strong basis for hypotheses about these relations
- Sampling biases are possible depending on how the cases (e.g., depressed clients) were identified and whether some special additional characteristic (e.g., coming to a clinic) was required
10
Q
Cohort Designs
A
- Cohort designs refer to strategies in which an intact group is studied over time, i.e., prospectively
- A cohort is a group of people who share a characteristic such as being born during a defined period of time
- In a cohort design, a group is followed over time-also referred to as a prospective, longitudinal study – to identify factors leading to an outcome of interest
- The group is assessed before the outcome (e.g., depression) has occurred. In contrast, in case-control designs, the groups (cases and controls) are selected based on an outcome that has already occurred
- The strength of cohort designs lies in establishing the relations between antecedent events and outcomes
- The time frame may vary depending on the goals of the study. In such a study, the antecedent condition is assessed (e.g., birth defects, early attachment) and once the outcome has not yet occurred (e.g., school competence, anxiety disorder). That is, the temporal order of antecedent and outcome is clear
- A single-group, cohort design identifies subjects who meet a particular criterion (e.g., children exposed to domestic violence) and follows them over time
- The basic requirements include assessment at least at two different points and a sample that, during that span of time, can change status on the outcome of interest (e.g., following survivors of 9-11 attacks overtime to examine who continues to show PTSD)
11
Q
Birth Cohort Designs
A
- There usually is a specific time frame and geographical locale
- Children are followed for an extended period – (sometimes) spanning decades
- The strength of the study derives from comprehensive assessments repeatedly over extended periods
o Participants are called back, usually for multiple assessments
o A large percentage of the sample must be retained; otherwise, selection-biases and poor external validity might result - Design enables understanding how development unfolds, in identifying the multiple paths toward adaptive and maladaptive functioning
- Effort, costs, and obstacles (e.g., retraining investigators, cases, and grant support) make birth-cohort studies are relatively rare
- “new” researchers usually are needed. The participants (infants) are likely to outlive the careers of the investigators who started the project
12
Q
Multigroup Cohort Design
A
- A prospective study in which two (or more) groups are identified (Time 1) and followed over time to examine outcomes of interest
- A two-cohort design begins by selecting groups that vary in exposure to some condition of interest (e.g., soldiers returning from combat) and follows them to see what the outcomes will be
- Distinct from a case-control design in that cases are followed prospectively to see what happens (i.e., the outcomes that emerge)
13
Q
Two-Group Cohort Design Example
- Does head injury in childhood increase the chances of later psychiatric disorder
A
- Youths who received head injury (e.g., accident) were identified and assessed over time for a 2-year period
- The obvious control group would be a sample of youths without a head injury, matched on various subject (sec, age, ethnicity) and demographic variables (e.g., social class) that are known to influence patterns of psychiatric disorders. However, a noninjury group may not provide the best comparison
- Perhaps any injury (whether to the head or toes) that leads to hospitalization for a child (or anyone) is traumatic and that trauma and entry into a hospital alone could increase later impairment
- In this study, the second group (making it a two- or multi- group cohort design) consisted of youths who were hospitalized for orthopedic injury (e.g., broken bones from accidents)
- Thus, both groups experienced injury, but head injury was the unique feature of the index group expected to predict later psychiatric disorder. Both groups were followed for 2 years after the injury and evaluated at that time
- As predicted, the results indicated that youths with head injury had a much higher rate of psychiatric disorder at the follow-up 2 years later when compared with orthopedic injury youths
- Potential problems in interpretation?
14
Q
Accelerated, Multi-Cohort Longitudinal Design
A
- A prospective, longitudinal study in which multiple groups (two or more cohorts) are studied in a special way
- Inclusion of cohorts who vary in age when they enter the study
- The design is referred to as accelerated because the period of interest is studied in a way that requires less time than if a single group were followed over time
- This is accomplished by including several groups, each of which covers only a portion of the total time frame of interest. The groups overlap in ways that permit the investigator to discuss the entire development period
- The design can identify if the characteristics of a particular cohort are due to historical influences in which the cohort is assessed
- In a single-group cohort design, a group is followed over an extended period and it is possible that the information generated by the group is special in light of the period in time in which the study was completed
- Main advantage is addressing the length of time it takes to complete longitudinal study
15
Q
Strengths of Cohort Designs
A
- Can establish the timeline (antecedent becomes before some outcome of interest)
- Measurement of the antecedents are not be biased by the outcome (e.g., being depressed now could not influence past recall of events)
- Multiple assessments at different points in time can be used to assess the predictors to chart the course or progression from the antecedent to the outcome
- All of the permutations can be studied in relation to the antecedent (occurred or did not occur at Time 1) and outcome (subjects did show or did not show the outcome at Time 2)
- Good for generating and testing theory about risk, protective, and causal factors and mediators and moderators
16
Q
Weaknesses of Cohort Designs
A
- Prospective studies can take considerable time to complete, and answers to critical questions (e.g., effect of asbestos on health, effect of abuse on youths) may have delayed answers
- Studied conducted over time can be costly in terms of personnel and resources. Retaining cases in a longitudinal study often challenging
- Attrition or loss of subjects over time can bias the sample
- Cohort effects may serve as a moderator i.e., it is possible that the findings are due to the sample assessed at a particular point in time
- The outcome of interest may have a relatively low base rate. Statistical power and sample sizes become issues to evaluate the outcome
17
Q
Levels of Specificity of the Construct
A
- Broad variables (e.g., age, sex, SES) are less preferred as the basis of an investigation than specific variables with which these may be associated (e.g., patterns of interacting with friends, social support patterns)
- Specifics constructs help move from description (e.g., males and females differ) toward explanation (e.g., processes that may explain the differences)
- Broad constructs often serve as a useful starting point at the beginning of research. However, understanding is enhanced by moving toward more specific con structs that might explain the processes through which the outcome might occur
- When two or more groups are compared (e.g., depressed vs. not depressed), operationalizing the criteria to delineate groups is critical
- The use of a single method of assessing a characteristic may restrict generality of the conclusions. For example, self-report of alcohol consumption may yield different results from other report or direct observation in the lab
- Regardless of what measure or operational criterion is invoked to classify subjects as cases or controls, we want to be sure that the measure is consistent and accurate in how individuals are classified
18
Q
Some Considerations for Specifying the Construct
A
- How are the groups delineated? On what measure?
- Usually rely on a particular measure (e.g., diagnostic instrument or a particular scale or questionnaire)
- Sometimes only the extremes of the distribution are considered (e.g., greater than the 67th percentile vs. less than the 33rd percentile). The rationale is that it is the middle group that is likely to be more unreliably identified
- Selecting extreme groups can be useful but deleting a large segment of the sample (e.g., the middle third) can distort the relations among the measures
- The continuum can be divided (e.g., high, medium, low) on some characteristic for purposes of description but the full range to see the relation of one variable (e.g., depression) with another (e.g., later eating disorder)
