Lecture 15: the correlational research strategy (part 2)

what are the timepoint based correlations? (3)

• cross-sectional: tests whether 2 variables measured at the same time are related to each other

• cross-lag: tests whether a variable from an earlier time point is associated with a variable at a later time point  

• autocorrelations: tests whether a single variable at one time point is related to the same variable at another time point 

what do we measure with timepoint based correlations? (3)

• strength

• direction: positive or negative

• significance

define “cross-sectional correlations”

tests whether 2 variables measured at the same time point are related to each other

true or false: the assumption of independence in a scatterplot (one point = one person) is met in cross-sectional correlations 

false: it isn’t because you measure the same people over time 

define “lag cross-correlations” (AKA temporal precedence)

tests whether a variable from an earlier timepoint is associated with another variable at a later timepoint

⚠︎ variable change can go both ways: X to Y and Y to X

what’s an assumption of the lag cross-correlation?

the lag value stays the same at each measurement

lag = […] -[…] (2)

• lag = time X - time Y

• lag = time Y- time X

why can we compute the lag by doing time X - time Y and by doing time Y - time X?

• we want to look at the positive (X - Y) and negative (Y - X) correlations to find what truly causes the lag

• we consider what happens what happens if X happens before Y and what happens if Y happens before X

define “autocorrelations”

test whether a single variable at one time point is related to the same variable at another time point

what is the characteristic that indicates that you can use autocorrelations?

something that is cyclical (like circadian cycles)

⚠︎ no need to, it’s just an indicator 

true or false: just like in lag cross-correlations, autocorrelations also consider positive and negative (auto)correlations

true: (but the lag value stays the same for lag cross)

• positive autocorrelation: lag increases over time

• negative autocorrelation: lag decreases over time

if the timepoint based correlations are significant, what does it mean? (explain individually for all 3 of them)

• cross-sectional: one variable covaries with the other variable

• lagged cross-sectional: one variable has temporal precedence over the other variable

• autocorrelation: one variable shows regular/repeating change over time

what’s the weakness of timepoint based correlations?

it doesn’t indicate causality, it only tells you relationship and strength (just like basic correlation)