causal relationship
= one variable directly/ indirectly influences another one (changes in values of one variable cause directly/ indirectly changes in other)
- can be unidirectional (variable A influences B but NOT vice versa; A–> B)
- bidirectional (each variable influences the other A–> must be able to demonstrate that variation in one of observed variables is only due to the influence of other observed variable
correlational relationship
= two variables accompany together but no test to assume causality
correlational research
- determine wether 2 (or more) variables convey and if, establish directions, magnitudes, form of observed relationship
- only observing them as is (no attempt to manipulate variables) –> no experimental research
- if correlational relationship exists > possible to predict from value of one variable the probable value of other one
- -> predictor variable (used to predict)
- -> criterion variable (being predicted)
- wether linkage between these variables is causal remains open question
third-variable problem (=lurking variable)
= possibility that correlational relationship may result from the action of an unobserved third variable, that may influence both of observed variables, causing them to vary together even through no direct relationship exists between them
directionality problem
= even when direct causal relationship exists, the direction of causality is sometimes difficult to determine;
–> challenge of determining which factor causes which
why use correlational research (3 situations)
- identify potential causal relationships during early stage of experiment can provide rich source of hypothesis that later may be tested experimental
experimental research
- high degree of control over variables you study
- if used properly, permits you to establish causal relationships among variables
- -> manipulating one ore more variables
- -> control over extraneous variables
independent variable
= values are chosen & set by the experimenter (values of independent variable is independent of participants behavior)
- these set values –> levels of independent variables (must expose participants at least to 2 levels of that variable)
- specific conditions associated with each level –> treatments
- -> by manipulating hope to show changes in level of independent variable –> causes change in behavior being reported
dependent variable (dependent measure)
= value that you observe & measure
experimental group
= group receiving/ exposed to experimental treatments
control group
= group that gets tested in exact same way as experimental group except that they get not exposed to experimental treatment what should be tested
- provides baseline of behavior, to which experimental group is compared
extraneous variables
= may effect behavior you wish to investigate but are not of interest for the present experiment
- can produce uncontrolled changes on value of dependent variable
- -> may make it difficult/ impossible to detect effects of independent variable
- -> may produce chance differences in behavior across the levels of independent variables
- -> could make it appear as through independent variable produced effects when it didn’t
control effect of extraneous variables (2 ways)
- hold extraneous variable constant (make sure all treatments are exactly alike, except for level of independent variable)
- randomize their effect across treatments (e.g. random assignment)
demonstration
= just shows what happens under specific conditions (not show causal relationship)
–> lack of crucial features: independent variable –> exposes group of subjects to only ONE treatment condition
quasi-independent variable
= correlational variable that resembles an independent variable in an experiment; created by assigning subjects to groups according to some characteristics they possess (e.g age, gender, IQ), rather than using random assignment
cross-sectional design
= creating groups based on the chronological age of your participants at the time of the study (several participants from each of a number of age groups)
–> permits you to obtain useful developmental data in relatively short period (don’t have to follow same participants over several years)
generation effect
= influence of generation difference in experiment (different generations for in different decades –> alternative explanation for observed difference)
–> threats internal validity
longitudinal design
= single group of participants is followed over some time period (alternative to cross-sectional design)
- the longer the period, the more difficult to keep track of participants
problems:
- cross-generational effects (results
might not apply to a different generation)
- subject mortality (if nonrandom)
- carry over effects
- history
cross-generation effect
= conclusion drawn of a certain generation may not apply to another generation
(generation effect in longitudinal design)
subject mortality
= loss of subject fro research over time
- problem if participants drop out of research because of nature of study –> creates a bias sample
- maybe special qualities of remaining participants –> results may not apply to general society
simpson’s paradox
= reversal of direction when association or comparison that holds all of several groups, data is combined to form a single group
- extreme form of the fact that observed observations can be misleading when there are lurking variables
subgroup of third-variable problem
- moderator
- mediator
- confounding variable
- = just influences the relationship of two variables; determines how strong relationship between A & B is
- influences independent & dependent variable, there might not even be a correlation between these two
- = influences just your independent variable
developmental design
= establish relationship between change in behavior and chronicle age
- -> age can’t be randomly assigned
- -> therefor, design is considered to be correlational or as having a quasi-independent variable
subgroups: cross-sectional/generational design; longitudinal design
