Chapter 7: Correlational and Differential Methods of Research

What 3 possibilities exist if A and B are correlated?

(1) A causes B (2) B causes A (3) some third factor C, causes both A and B (although we don't specify what it is - could be hundreds) In the abstract, any of these possibilities could be true. However, in real life, one or more of the possibilities might appear implausible, leading the researcher to draw a strong casual conclusion. Also should consider what the relationship means in operational terms: how do we quantify the relationship? What do the scores on a test testing A and B really mean (one variable might me important regardless of the other)?

An artifact

- Any apparent effect of an independent variable that is actually the result of some other variable that was not properly controlled - The result of confounding

Differential research methods

- Compares two or more groups that differ on preexisting variables (differences existed before the researcher conducted the study) - Qualitative dimensions (gender, political party affiliation, or psychiatric diagnosis) or quantitative dimensions (participant's age or number of years of education) can define the group - Classification variable is the independent variable (non-manipulated independent variable). Behaviors measured in the different groups is the dependent variable. - Avoid drawing casual conclusions because variables are not manipulated - Same statistical proceedures used to evaluate the data as from experimental methods (because both use different groups defined by an independent variable & the researcher measures a dependent variable on all participants in each group)

Longitudinal designs

- Follow the same people over time to observe developmental changes (controls for cohort effects) Disadvantage: take a long time to complete

What are the advantages of low-constraint research?

- It's flexible and allows us to explore a phenomenon, gaining a sense of what to expect. Allows us to design appropriate procedures for testing our hypotheses. Scientists, however, may not need to start with low-constraint research because they often study what others have already studied extensively.

Measuring variables in differential research

- Measure, rather than manipulate, the independent variable in differential research Need to decide on the operational definition of the independent variable (set of procedures for data collection) Although non manipulated independent variables in differential research are usually discrete variables, it is always possible to take a continuous variable and break it into discrete intervals (ex: education level into high school dropout, high school graduate, etc.). Converts a correlational research design into a differential research design.

Filler items

- Measures not meant to measure anything, but intended to draw the participant's attention away from the real purpose of the measure

What does measurement depend on?

- The adequacy of operational definitions

Correlational Research Methods

- Quantifies the strength and direction of the relationship between (at least) two measures. Not the same as causality (can't prove a theory). - Don't manipulate variables - Plans for measuring variables are formalized prior to measurement

Selecting appropriate control groups in differential research

- Researcher must decide which groups to include a differential study Control group and experimental group

Sampling

- Researchers must sample randomly from the population to generalize to a larger population. Usually researcher selects a sample from all the available participants. - Random sampling: procedure for selecting participants from a population so that each participant has an equal change of being selected - Threat to generalizability is the subtle bias that can occur when a researcher has access only to certain groups (ex: examining Alzheimer's from one nursing facility - facilities specialize in kinds of individuals they treat, specific economic class and education level) - Even when researchers appear to be sampling randomly, it is important to be sensitive to subtle biases (ex: location, time of day or day of the week at a particular location - could create a sample that doesn't represent the overall population) - Also researchers might approach people who seem more likely to cooperate and avoid those who seem in a hurry.

Correlational research

- Seeks to quantify the direction & strength of a relationship among two or more variables Involves 1. Problem statements: specific (unlike flexible ones in naturalistic research and case studies). "What is the strength and direction of the relationship between variable X and variable y?" "What is the best equation for predicting variable Y from variable X?" Regression equation 2. Secondary Analyses: helps to explain findings. Ex: use demographic variables (characteristics of individuals such as age, education and social class) to see if these variables could produce artifacts which would explain the correlations of the study. Ask: "What is the correlation of each of the demographic variables with the dependent variable(s)?" Compute these correlations separately within each group because the relationship might be different in the groups.

Problem statements for differential research

- The most challenging in all of research, however, on the surface appear to be straightforward Ex: "Does group A differ from group B on the dependent variable(s)?" Must make comparisons that have theoretical significance - pick the right groups & the right dependent variable to advance scientific understanding A theoretically significant differential research study will tell us something about factors that affect the dependent variable, rather than just revealing differences between groups Rules of thumb for choosing problem statements: (1) Should develop problem statements that focus on comparing groups that differ on only one variable ex: comparing fifth grade girls to fifth grade boys to understand gender differences, rather than fifth grade girls to college professors (2) Use several comparisons when trying to draw a conclusion about the effects of a factor from differential research studies ex: to examine gender differences should look at fifth grade girls and boys, college males and females, male and female business people, etc. Even from this should be careful in drawing casual conclusions (3) Focus group differences on theoretically relevant dependent measures ex: for studying self-esteem would focus on variables like success, social support, childhood experiences and cognitions instead of hat size, finger tapping speed or the make and model of one's car for example

Confounded

- Two variables are confounded when they vary at the same time (fail to keep one variable constant, while we vary the other). Makes it unclear which variable is responsible for observed differences in the dependent variable. Ex: As the group variable changes, the method of observation also changes - To avoid confounding variables, must hold constant the variable of least interest and let the variable of greatest interest vary For differential this means holding the observational method constant & allowing the group variable to vary. Failure to constrain procedures can lead to confounding and artifacts.

Differential Research Methods

- Used to compare existing groups when experimental procedures for testing group differences cannot be used for practical or ethical reasons Includes problem statements, measuring the variables, selecting appropriate control groups and sampling

Another design that falls between cross-sectional and longitudinal design

- Uses successive independent samples over time, asks the same questions with each testing, but asks them of a new sample of participants

Time-series designs

- Variations of longitudinal designs that involve multiple measurements taken before and after a manipulation Can be used with individuals or groups

When is differential research design used?

- When the manipulation of an independent variable is impractical (too expensive and complex), impossible (independent variable can not be manipulated) or inappropriate (unethical)

How is establishing a correlation useful?

1. Correlations can be used to predict future events, even if we have no idea of why the relationship exists. ex: Ptolemy (A.D. 140) model of planetary movement accurate, although that they revolved around the Earth 2. To provide data either consistent or inconsistent with scientific theories -> a correlational study cannot prove a theory correct (it can make us more confident in the theory), but it can negate a theory

What is a moderator variable?

A variable that seems to modify the relationship between other variables Like gender, culture or ethnicity For ethnicity we now have cross-cultural research (culture can significantly influence many psychological variables). Should compute correlations for different subgroups to increase confidence that the relationship will hold for the entire population

What are the differences between correlational and naturalistic?

In correlational, - always measure at least two variables - our plans for measuring variables are formalized prior to measurement

What do some theorists disagree with?

The statement that correlations do not imply causality Some researchers argue that the right combination of correlational studies can so effectively exclude other interpretations of a complex data set that a casual interpretation is reasonable However, research literature full of examples of top scientists drawing what later proved to be incorrect casual interpretations from complex correlational data

Experimental group

The primary group

Comparing correlational and differential research methods

- Both measure relationships among variables. - However, differential is higher-constraint because differential is interested in casual questions that constraint experimental research ethically or practically (ex: impossible to assign participants randomly to a group of people with schizophrenia & a group without). - Yet, differential research often limited by confounding variables (pre-existing groups include groups of people who differ on several other variables than the one we are measuring) - To control confounding variables, in order to draw strong conclusions, must select a general control group that is comparable to the experimental group on one or more potential confounding variables (ex: if we think schizophrenics are often of a lower social class affecting the dependent variable, we would find a control group of people of lower class). Active control over sampling.

What is a field that often uses correlational and differential research designs due to ethical considerations?

- Clinical neuropsychology Uses measures of behavior to infer the structural and functional condition of the brain Ex: identifies relationships between behavior and brain dysfunction

What other than sampling can affect the generalizability of a study in differential research?

- Number of participants who drop out - People with physical or psychiatric disorders might not have the concentration to complete a long and demanding study People who perform the task and different from the overall population

What is a major concern in research?

- Obtaining a sample that adequately represents the population to which the researcher wants to generalize Important in correlational research..... - Is the relationship between a given pair of variables the same in all segments of the population. If researcher suspects that such differences exist (like moderator variables - gender, culture and ethnicity), they might draw samples from separate subpopulations.

Cross-sectional design

- Part of a differential research design because participants assigned to groups based on the preexisting characteristic of age - Groups of participants at different ages are compared on a set of variables - Need to be cautious in drawing conclusions Cohort effect: the shared life experience of people of a given age in a given culture may lead them to behave similarly throughout their lives, but differently from people of other ages or cultures (ex: people who lived through the Great Depression often have similar thinking, expectations and emotional responses from this powerful experience -> when looking at the aging process, shouldn't connect being cautious about going into debt with the aging process instead with the GD).

What does higher-constraint research require?

- Precise and consistent observational procedures and detailed planning Causes us to lose flexibility Seldom used in the early stages of studying a problem because detailed planning is impossible when we don't have a reasonable understanding of the phenomenon under study

Control group

Any group selected in differential research as a basis of comparison with the experimental group Control this group to reduce the effects of potential confounding variable. Variable can have a confounding effect if it... (1) affects the scores on the dependent variable(s) (2) there is a difference between the experimental and control groups on the potential confounding variable The ideal control group is identical to the experimental group on all variables except the independent variable that defines the groups (ex: effect of exposure to toxic chemicals on cognitive performance - experimental group exposed & control group non-exposed, but control group is same age, social class, education level, similar kind of work). Rare to find an ideal control group - researchers usually try to find a group that controls some of the most important and powerful confounding variables (variables that are likely to have a large effect on the dependent measure). Or researcher uses multiple control groups - each control group typically controls for one or more for the major confounding variables, but no group controls for all. Research often involves multiple studies by different researchers in different laboratories each using slightly different procedures and control groups, because it is not always feasible to include all possible comparison groups in one study.

How is confounding minimized?

By active control over sampling Correlational research has no comparable control procedure -> why differential research is higher constraint

How do researchers avoid the possibility of unintentionally influencing participants?

Control two effects (1) experimenter expectancy: the tendency of investigators to see what they expect to see Minimize this by using objective measures whenever possible. (2) experimenter reactivity: the tendency of investigators to influence the behavior of participants Minimize this by using two independent researchers or automating the research procedures so the researcher need not be present. Another potential problem is the participant's influence. Participants like to be consistent, especially when they know that researchers are observing and evaluating them = artificial consistency, giving the impression of s strong relationship between variables when no relationship or a weak relationship actually exists. Reduce this effect by including filler items OR by relying on one or more unobtrusive measures OR separating the measures from one another, time separation does not have to be large (taking measurements at different times or having different researchers take the measurements) ex: researchers tell participants they are involved in two short studies in a single session, when the second study is part of the first study (don't get this??!) OR use measures beyond the control of the participant (not self reports or behavioral observations)

What can ethical constraints limit?

Experimentation Can receive three important lessons from this: (1) Know that the ethics of research are of extreme importance (2) Learn the ethical concerns, danger and the corrections (3) Understand that your vigilance is constantly required because strong pressures in our society continue to try to weaken ethical constraints in order to satisfy other needs (like personal ambition, political goals and commercial gain -> fraud). Problem association with collaboration among scientists in different countries -> monitoring criteria is not standardized. Use low-constraint research as an alternative. Will not tell us about causality, but can provide much useful information.