Chapter 3 - Research Methods

Measurement Issues - Reliability

- A measure is reliable if it yields the same results every time it is used. - Can be assessed by test-retest reliability. - Internal consistency: how similar the answers for two similar questions are/ - Inter-rater reliability: degree to which two ratings of the same data are similar.

Research Design - Field experiments

- Also in natural setting, but introducing an independent variable and measuring the outcomes. - Example: therapy dogs influence on care, contrasting with the control group. - Advantage: still control of the IV; greater ecological validity than the lab, less than the naturalistic - middle ground. - Disadvantage: less control than in lab, logistic challenges.

Multivariate Correlational Design

- Analyzes the relationship between more than two variables. - Simultaneously evaluates the effects of many potentially important factors. - Enables researcher to use a set of variables to predict scores on another variable. - Types: a. Multiple regression analysis b. Logistic regression c. Structural equation modelling (SEM) d. Hierarchical linear modelling (HLM)

Non-random Sampling

- Because of attrition, after a while the sample may not be representative of the general population anymore. - Researchers must keep track of the remaining group at different points to ensure it is still representative.

How to best control for cohort effects in cross-sectional studies?

- By selecting younger samples that are comparable in important ways to the older sample. - For example, selecting participants with similar professional or educational backgrounds.

Research Design - Correlational studies

- Correlational studies: relationship between two variables (NOT causality). - Positive correlation: variables move together in the same direction (+1) - Negative correlation: variables change in opposite directions (-1) - ZERO = no relationship, random. - Example: age and memory. - Age is a very good variable. Influences: change over time OR cohort effect (trend within a generation).

Terminal decline

- Decline to cognitive abilities that happens closer to time of death. - Point at which individuals can no longer participate in a study. - Because they were still in the sample as they declined, they may have pulled down the group's average. - Once they die, the researcher may think that the outliving group improved when in fact they just "got rid of" the weakest links.

Mos Efficient Design

- Developed by K. Warner Schaie (1965) - A set of three designs manipulating variables of AGE, COHORT and TIME OF MEASUREMENT. - Enables the most amount of information to be condensed into the most inclusive data framework. - Three designs: a. Time-sequential (time with age) b. Cohort-sequential (cohort with age) c. Cross-sequential (cohort with time) - Depending on the pattern of significant effects, it may be possible to draw conclusions about the relative influence of personal and historical aging on test performance.

Cohort

- Group with a shared characteristic. - In developmental psychology, that is the year or period of a person's birth.

Measurement Issues - Construct Validity

- Indicates if a measure of a psychological construct is actually measuring the construct. - Needs convergent and divergent evidence. - Convergent: determines if the measure relates to other measures that are theoretically similar. - Divergent: ensures the measure does not relate to other measures that have no theoretical relationship to it.

Multivariate Correlational Design - Hierarchical linear modelling (HLM)

- Individual patterns of change are examined over time, rather than simply comparing mean scores of people at different ages. - Particularly important for longitudinal studies, because not every participant exhibits the same patterns of change over time.

Research Design - Epidemiological

- May use surveys, interviews, biological samples, etc. - Study the distribution and determinants of health-related states or events. - Results provide Prevalence and Incidence statistics. - Prevalence: percentage of people who ever had symptoms over a period. - Incidence: percentage of people who first have symptoms in a given period.

Research Design - Naturalistic observation

- Natural setting, not controlled. - Watch for changes occurring naturally. - May create behavioural records and note how often behaviours are observed. - Examples: long term care facilities, workplace, home. - Advantage: ecological validity comes from the lack of control of the researcher. - Disadvantage: no control makes it difficult to make causation statements; time consuming.

Longitudinal Study

- One age group, tested more than once. - Interval can be of any length. - Biggest problem: tests must remain the same, so no flexibility for adopting new technology. - Practice effects: participants get used to the tests and "get better" at them (confound). - Attempt to determine whether changes have been the result of the aging process: participants are their own control group.

Research Design - Lab experiments

- Participants are tested in a systematic fashion using standardized procedures. - Less concern for confounding variables. - Very objective. - Advantage: full control - Disadvantage: less ecological validity, may not reflect the person's real ability due to individual reactions to the controlled environment.

Selective attrition

- Participants that choose to leave a study. - The tendency of some people to be more likely to drop out of a study than others. - Must test if the drop-outs share any characteristic that could affect the results gathered from the remaining participants: was the drop out pattern random, or is there a systematic bias at play?

Multivariate Correlational Design - Multiple regression analysis

- Predictor variable is regarded as the "independent" variable. - Predicted variable is regarded as the "dependent" variable. - Enables to suggest and test different inferences about cause-effect relationships.

Research Design - Others

- Qualitative: interviews, open-ended investigation - Case studies: focus on 1-2 people, gather info from different sources on the same person, relies heavily on the analysis of the researcher. - Archival: existing data, disadvantages include lack of control over data and incomplete data. - Survey: questionnaires or online, advantage of pooling more people than it would be possible in the lab. - Epidemiological: disease and disease progression. - Meta-analysis: results from multiple studies being pulled together, studies are chosen based on specific criteria. - Focus groups: meeting of a group to discuss a topic, researcher tries to identify themes in the discussion; goal is to develop research questions to pursue in later studies.

Time of measurement effects on individual performance.

- Related to any current situation or event, even if far removed from the individual, at the time they are brought in to be tested. - Even world events such as the European immigration crisis must be considered.

Correlational Design

- Researcher makes no attempt to divide participants into groups or to manipulate variables. - Simply looks at relationships between factors as they exist in the world. - There is no assumption of causation, no independent and dependent variable. - Two types: a. Simple, bivariate correlational: two variables. b. Multivariate correlational: more than two variables.

Research Design - Quasi-experimental

- Researchers compare groups on predetermined characteristics. - Placement of participants in the experimental and control group is NOT random. - Any study of aging is quasi-experimental because age can't be randomly assigned. - Experimenters will try to eliminate other variables but can only go so far as to infer a relationship between age and any result they get.

Longitudinal Design - Prospective Study

- Researchers take a sample of a population and wait for the variable they want to test to come through - wait for it to happen. - Example: research on widowhood may sample married couples and wait to study the surviving partner later in their lives.

Cohort effect

- Something that happened to that group specifically that affected their development, making them different from other groups. - In developmental research: refers to the social, historical and cultural influences that affected people during a particular period of time and history. - Related to Normative History-Graded influences.

Latent Variable

- Statistical composite of several variables that were actually measured. - Example: researchers can be measuring life satisfaction according to three different measures. They can construct a latent variables that then encompasses all three, like an "umbrella" variable.

Mediation

- Statistical test in which correlation between two variables is compared with and without their joint correlation to a third variable. - Example: comparing correlation between educational and life satisfaction and their joint correlation to health status. If the correlation between life satisfaction and education becomes smaller when health status is included, then health is serving the "mediator" role.

Moderation

- Test for the impact two variables seem to have on a third. - Example: both age and education may impact life satisfaction, but age and education aren't linked.

Path Analysis

- Testing all possible correlations among a set of variables to see if they can be explained by a single model. - Basically, tracing the path between the variables, who is affecting who, in what directions. - Paths are determined in terms of moderation and mediation. - Example: testing whether education and age both affect life satisfaction (moderation) through their joint effect on health (mediation), which in turn affects life satisfaction.

Multivariate Correlational Design - Structural equation modelling (SEM)

- Testing models involving relationships that include latent variables. - Similar to Path Analysis, they propose a set of relationships among the variables but some are directly measured while some as constructed as latent variables.

Multivariate Correlational Design - Logistic regression

- Testing the likelihood of an individual receiving a score on a discrete yes-no variable. - Example: testing the probability of a person getting a diagnosis of diabetes or nor, depending on a specific risk factor.

Measurement Issues - Validity

- The test measures what it is supposed to measure. - Content validity: indicates if a test designed to assess factual material accurately measures the material. - Criterion validity: indicates if a test score accurately predicts performance on an indicator. - Construct validity: indicates if a measure of a psychological construct is actually doing so.

Cross-sectional

- Various age groups, tested once. - No concern of attrition. - Practice effect is low but can still happen. - More able to pick up on cohort effects: must do more rigorous control of cohort effects. - Problem: usually the younger sample has a shorter difference between ages than the older sample. (e.g. they may test adults between 20 and 25 and compare them to adults between 50 and 70). - Problem: different age groups will react differently to the test materials. Older adults may be less familiar with test-taking, which can affect their scores.

Sequential

- Various age groups, tested various times. - Basically two or more longitudinal studies running together. - Also good for picking up on cohort effects. - Least popular, often labelled as longitudinal.

Three factors influencing individual performance on a given psychological measure?

1. Age 2. Cohort 3. Time of measurement

Advantages of Multiple Correlational Design (3)

1. Control for confounds related to age. 2. Allow investigations of "causality" 3. Provide ways to examine change over time

Types of research methods in development studies? (5)

1. Naturalistic observation 2. Correlational studies 3. Field experiments 4. Lab experiments 5. Others: qualitative, case studies, archival studies, surveys, epidemiological, meta-analysis

In developmental research, age can be said to "cause" specific differences between age groups. True or False?

False. Age is not a truly independent variable, so it can never be said to cause any differences between groups.

Correlation = ZERO

No relationship between the variables exists.

Negative Correlation

Variables increase or decrease in opposite directions.

Positive Correlation

Variables increase or decrease in the same direction.