Lecture 6 Flashcards
how to interpret a single error bar
> 95% confidence interval
95% confidence interval
NOT: if i would replicate the procedure 100 times, 95 sample means would occur within original 95% CI
BUT: if i replicate the procedure 100 times, I expect 95 of the 95% CI would contain mu
error bar of 1 SEM: how big is the confidence interval?
1 SEM error bar: 67% confidence interval
assessing significance in graphs
(between subjects design)
> rule of thumb for SEM
> role of thumb for 95% CI
rule of thumb for SEM:
> significant if minimally 1 SEM fits in between the error bars of both groups
rule of thumb for 95% CI:
> significant if the overlap of the error bars of both groups is smaller than 25% of the length of both CI
why do we need to use different rules of thumb for between/within subject designs?
the error term in within subjects designs derives from the participants x factor interaction
> no fixed relationship between error bars and interaction
relational research
> no dependent / independent variable, but…?
relational resarch:
> predictor and criterion
what are the two main measures in relational research?
> what kind of data?
- pearsons product -moment correlation coefficient r
> between two continuous, normally distributed variables
- spearmans rank-order correlation coefficient rs
> between two ordinal variables
you observe r(x,y) = 0
> what can you conclude`?
r(x,y) = 0
> no relationship between x and y
> insufficient power i.e. type 2 error
> truncated range prevents the expression of a relationship
> the relationship between x and y is not linear
you observe r(x,y,) =! 0
> what can you conclude
r(x,y) =! 0
> there is a linear relationship between x and y in the population
> type 1 error
> an outlier caused the relationship
what is the use of spearmans rank order correlation coefficient?
spearmans rank order correlation coefficient
> using rank ordered scores, rS is more resistant to outliers
regression model:
how does SStotal look as a regression line
SS total:
> horizontal regression line at Ymean
> the sum of the squared deviations of each data point to Ymean
regression model:
> how does SSerror look like as a regression line
SSerror
> deviations from the actual regression line that fits the data best
> also called SS residual
how can you describe the amount of explained variance using SS?
proportion variance explained = r²
r² = (SStotal - SSresidual) / SStotal
under which circumstances is it possible to prove causality using linear regression?
general issues of experimental research
+ manipulation independent variable
