ch4 internal validity

constant B

all the Ps in the study have the same value on that B/measure

variable B

at least some of the Ps have different values on that B/measure

measured B

the value of that B/measure is obtained by observation or self-report of the P

manipulated B

the value of that B/measure is controlled, delivered, etc. by the researcher

causal RH

the value of the variable manipulated by the researcher causes the value of the variable measured from the P, must have temporal precedence, a demonstrated statistical relationship, and elimination of alternative explanations

confounds

any variable other than the one manipulated by the researcher that systematically co-varies with the causal variable, so we can’t logically determine which variable is at work

control constants

any B for which all Ps have the same value, ie. all Ps are 6 years old

control variables

any B for which Ps have different values, but for which the treatment/conditions are “balanced” or “equivalent” on that variable, ie. half the Ps in a study are male and half are female

initial equivalence

prior to manipulation of the causal V, Ps in different conditions are the same on all meaasured Vs

ongoing equivalence

during manipulation of the causal V and completion of the task and measurement of the effect V, Ps in different conditions are the same on all manipulated Vs except the causal V

how do we produce initial equivalence

random assignment

how do we produce ongoing equivalence

procedural standardization of manipulation, confound control, task completion, and performance measurement (there’s no cure for confounds, they are only avoided by knowledge of their existence)

between groups design

each P will complete only one condition, RA determines which condition for each P

within-groups design

each P will complete all conditions, RA determines the condition order for each P

random assignment of groups (rather than individuals)

unacceptable assignment where groups are unknowingly equivalent (like if a friend group signs up together)

arbitrary assignment by the researcher

unacceptable assignment where a non-probabilistic process is used (could be based on a good idea but isn’t random)

self assignment by the P

unacceptable assignment where P chooses what condition/order they will be in

administrative assignment

unacceptable assignment where non-random assignment is determined by someone other than the researcher

non assignment or natural assignment

unacceptable assignment where P is already “in” conditions before they arrive at the study

2 things to look for to help evaluate ongoing equivalence

1. research conducted in a lab

2. shorter procedure

trade off characterization

it’s impossible to promote both internal and external validity within a single study, so the researcher must choose which to emphasize (internal validity/control or external validity/representativeness)

precursor characterization

focuses on causal information, suggesting associative info is not valuable