Ch 2 Internal and External Validity

Experimental methods

these methods involve research in which participant actions are limited or in some way constrained by the controlled manipulation of variables determined by the researcher

quasi-experimental methods

these are a variant of experimental methods lacking in random assignment, but participants are exposed to some form of variable manipulation imposed by the researcher

nonexperimental methods

on these methods no variables are manipulated. Instead, the relationships of naturally occurring variables are measured by the researcher

Unidirectional causation

a relationship in which changes in A produce subsequent changes in B, but changes in B do not influence A (e.g., increases in the temperature-humidity index produces an increase in aggressive responses of rats, but the degree of aggressiveness of rats does not affect weather conditions)

Bidirectional causation

a relationship in which changes in A produces changes in B and, in addition, changes in B produce changes in A (e.g., perceiving threat produces feelings of anxiety, and increasing anxiety enhances the perception of threat)

Noncausal covariation (correlation or third-variable causation)

a relationship in which changes in A are indirectly accompanied by changes in B, because both A and B are determined by changes in another variable, C. For example, increases or decreases in the cost of living, C, results in the rise or fall of birth rate, A, and consumption of beefsteak, B.

moderator

a third variable that can either augment or block the presence of a predictive relationship

mediator

a third variable that serves as intermediary to help explicate (develop/further) the chain of process of a predictive relationship

independent variable

a characteristic manipulated independently of its natural sources of covariation to produce different conditions in an experimental study

dependent variable

measured outcome or consequence not manipulated by the researcher, and expected to be influenced by (dependent upon) manipulation of the independent variable

internal validity

the extent that inferences of causality could be made about the obtained relationship between the independent variable and the dependent variable

external validity

the extent of generalizability or certainty that results can be applied to other respondent groups, different settings, and different ways of operationalizing the conceptual variables

statistical significance

this is achieved if the probability of obtaining the observed effect by chance is so low as to render the chance explanation implausible. Probability of .05 or less

statistical power

the probability of obtaining a statistical significant effect, if indeed that effect truly exists.

confound

a type of extraneous variable in which its effect and the effect of the independent variable on the dependent variable cannot be separated; thereby it poses a threat to internal validity

instrumentation error

scores are caused by changes in the properties of the measurement instrument, rather than by changes in the participants being measured.

statistical regression

outcome scores are subject to misinterpretation if participants are selected on the basis of extreme scores at their initial measurement session

selection error

if scores for two or more groups of participants are being compared, differences are caused by selection (nonrandom) procedures employed when participants were assigned to groups

mortality

selection procedures, treatment differences, or issues that yield different proportions of participants dropping out of the study may cause the observed differences between the groups in final measurements

selection-based interactions

If participants were differentially (non-randomly) selected to serve in comparison groups, these specially selected groups may experience differences in history, maturation, testing, etc., which may produce differences in the final measurement.

random assignment

requires that all persons available for a particular research study be able to participate in either the experimental or control group, and that only chance determines the group to which any individual is assigned, increases study's internal validity

random selection

requires that chance determines selection of participants for a study, and is used to assure generalizability of results from the sample to the entire population of relevant persons, all persons in a population of interest are equally likely to be included in the research sample, increases study's external validity,