Reviewer 2

• Alternatives to experimentation include correlational, and quasi-experimental approaches.
• These methodologies are used when traditional experimental manipulation isn't feasible or ethical.

Objectives of Non-Experimental Techniques

• These techniques can be used to understand the interpretation of nonexperimental results
• Causal models can be constructed from correlation-based designs.
• Allows learning of techniques that do not manipulate antecedent conditions allowing the use of correlations, other correlational-based methods, and quasi-experimental designs

Quasi-Experiments

• "Quasi" in Latin translates to "seeming like."
• These closely resemble experiments but lack manipulation of antecedent conditions and/or random assignment.
• They study the effects of preexisting conditions like life events(9/11) or subject characteristics like autism on behavior
• For instance, comparing the incidence of Alzheimer's in patients who used ibuprofen since age 50 and those who did not.
• Use of quasi-experiments is recommended when antecedent conditions cannot or should not be manipulated
• An example; can be applied to the effects of spouse abuse on the frequency of child abuse.

Problems with Quasi-Experiments

• Causality cannot be established with certainty (cause and effect).
• Limited internal validity which is the ability to conclude with confidence that the antecedent condition caused observed behavioral differences
• Higher external validity, generalizability, when compared to lab experiments.
• Low manipulation of antecedents, but high in the imposition of units.

Correlations

• Correlations are used to establish relationships among pre-existing variables.
• Correlations show relationships between antecedents and behavioral effects, where antecedents are pre-existing, not manipulated.
• Correlation studies determine the degree of relationship between two traits, behaviors, or events.
• Changes in one factor are associated with changes in the other.
• Selected traits/behaviors measured and recorded, then degree of relationship determined statistically.
• Correlations allow the prediction of one set of behaviors from another.
• College grades can be predicted from high school grades, and vice versa.
• There is a low manipulation of antecedents, but high imposition of units.
• It Cannot be sure of cause
• A correlation study has poor internal validity but good external validity
• Once a correlation is known it can be used to predictions.
• If a person's known score on one variable, we can make a better prediction of that person's score on a highly correlated measure.
• The higher the correlation, the more accurate the prediction will be.
• Example: A researcher can study the relationship between watching Sesame Street and vocabulary in preschoolers.
• Parents are asked to list words known by preschoolers and how frequently they watch Sesame Street.
• One variable is hours per week, other variable is number of words the child knows.
• Numbers/data are statistically analyzed for all children in the study.

Calculating Correlations

• The Pearson Product Moment Correlation Coefficient (r) calculates a correlation.
• The correlation can be positive, negative or show no relationships.
• Correlation coefficient values range between -1.0 and +1.0.
• The "+" or "-" sign Indicates relationships are positive or negative.
• Absolute value of 'r' shows the relationship strength.

Scatterplot

• It demonstrates direction of a correlation and used for initial analysis.
• Each dot represents a person's scores, with each person having two scores.
• Regression lines, or lines of best fit, are drawn on scatterplots.
• Line direction signifies the direction of the correlation.

Positive Correlation

• A positive "r" value indicates a positive correlation between variables.
• As one variable increases, the other increases, and vice versa.
• More hours of Sesame Street increases vocabulary.
• Absolute "r" value shows relationship strength; the closer to 1.00, the stronger the correlation.
• Strong relationships allow for good prediction.

Negative Correlation

• A negative "r" value indicates a negative correlation between variables.
• As one variable increases, the other decreases.
• An example; increase in the hours of Sesame Street viewing decreases vocabulary.
• The absolute value of r tells how strong the relationship is
• The closer it is to 1.00, the stronger it is. Strong relationships allow for good prediction

No Relationship

• Absolute value of correlation close to 0, means no relationship between the variables
• For example, Sesame Street viewing no effect on vocabulary.

Curvilinear Relationships

• A correlation coefficient value near zero may incorrectly suggest no relationship.
• A curvilinear relationship between variables can be seen on the scatterplot.

Properties of Correlation Coefficients

• Linearity: shows the relationship between variables X and Y and can be plotted as a line(linear) or a curve (curvilinear).
• Sign: correlation coefficient is positive or negative.
• Magnitude: strength of the correlation coefficient, ranging from -1 to +1.
• Probability: the likelihood of obtaining a correlation coefficient of this magnitude due to chance.

Scatterplots

• Scatterplots a graphic display of pairs of data points on the x and y axes.
• Illustrates linearity, sign, magnitude, and probability of a correlation indirectly)

Outliers

• Outliers are extreme scores affecting correlations by disturbing data trends.

Correlation vs. Causation

• Correlation does not imply causation, even with a perfect correlation. Other variables not measured could influence the effects.
• Research on firmness firm handshakes and positive first impressions show a positive correlation
• Cannot be said whether the handshake or the personalty trait caused a good impression
• A positive correlation between the number of cars built and the number of airplanes built. Each is not the sole causation of each
• Calculate the coefficient of determination (r²) to estimate the amount of variability explained by a predictor variable.
• The coefficient of determination is an estimate of strength
• Correlational studies cannot establish a cause and effect relationships, because correlational studies do not create multiple levels of an independent variable and randomly assign subjects to conditions

Reasons Why Correlation Doesn't Prove Causation

• (1) Causal direction - uncertain that a cause and effect variable exists
• (2) Bidirectional causation — both variables could influence each other.
• (3) the third variable problem - There could be some other variable that is the cause that we have not measured
• Causal direction- Correlations are symmetrical, A may cause B as readily as B causes A.
• Bidirectional causation-Two variables like Insomnia and depression- may affect each other
• The third variable (family conflict) may create a third variable and that cause that both insomnia and depression are each other
• Linear regression analysis- researchers use linear regression analysis to estimate someone's score/behavior on another and there is strongly related relationship between the two
• Multiple regression- Researchers use this to to predict behavior from the other variable by measuring a variable that has to do with the behavior.

Quasi-Experimental Designs

• Lacks important elements of experiments, manipulates antecedents or random assignment to treatment conditions.
• This design doesn't look for relationships between variables; they compare groups or track changes over time in one group.
• These designs have low internal validity, we don't fully understand cause and effect. Ex post facto ("after the fact") design- examines existing, unmanipulated subject variables.
• For example, to study women who are divorced to see if they are more pessimistic about marriage than women who are not divorced and who are single. The preexisting variable or subject variable
• Correlations- there is no cause. So cannot say that divorce causes changes in attitudes. There could have been a third variable which caused the result.
• Ex post facto studies are low in internal validity.
• Nonequivalent groups design- compares effects of treatments on preexisting groups.
• Installs fluorescent lighting in Company A and incandescent lighting in Company B and assess productivity.
• Cannot be sure that the lighting made the difference; it could be that one company is threatening layoffs, so workers are being more diligent.Low internal validity
• Longitudinal designs measure the same subjects at different times to see time's effect on behavior.
• Whereas, cross-sectional studies compare subjects at different life stages at one point in time.
• Pretest/Posttest Design
• Researchers use pretest/posttest designs to measures behavior before and after an event. It is quasi-experimental because there is no control condition.
• Example: Practice GRE test 1 → six-week prepare course → Practice GRE test 2.
• There us no control group with a different level of independent variables.
• Results may be confounded by practice effects, caused due to less stress or learning of pretest answers People do better the second time they take an intelligence test, there is no special training in between.

Solomon 4-Group Design

• Variation on pretest/posttest design including conditions:
• A group with pretest, treatment, and post-test.
• A nonequivalent control group with pretest and posttest only.
• A group that received the treatment and a posttest
• A group that only received the posttest