Exploratory factor analysis Flashcards
What are principal components?
estimates components that account for 100% of total variance in variables
What are factor techniques?
estimates components that account for 100% of shared variance between variables
What is the least squares method?
minimises diff between data and factor analysis
Whar is the max likelihood test?
finds most probably factor analysis
What is dimensionality?
no. of variables or k
What are factors?
the r/s between 2 or more variables
What is an eigenvector?
the direction of the factor
What is an eigenvalue?
the amount of variance in the eigenvector
What are the requirements for factor analysis?
min 300 sample, 50 is poor, 1000 is excellent
What is monte carlo testing?
generating own distribution for factors
What is pairwise deletion of data?
deleting data that is missing
what is listwise deletion of missing data?
deleting a whole case if there’s missing data
What is multiple imputation?
creating sev data sets and replacing missing data with imputed values, all slightly diff in each data set due to random component, analyse them all and combine results, then calculating variation in estimates to report on it
Expectation maximisation?
finding max likelihood estimates for model parameters when data is incomplete
What is regression replacement?
putting known ave. for variable in place, however this reduces variance in data
What is multivariate normality?
normality across all r/s in data set
What is communality?
how much variance is explaned by variables
What is in the factor matrix?
b-weights for 1 unit changes in ivs
What sampling techniques to use to check if sample is adequate?
kaiser-meyer, has to be greater than .5, and bartlett’s less than .1
When to stop extracting?
when you get to k, eigenvalue>1, when the slope changes dramatically
Residual correlation matrix
1-.20 not acceptable, .02-.05 evidence of unexplained factor, .05-0 no evidence
What is quartimax rotation?
reduce no.of variables to explain each variable, simplifies
what is a varimax rotation?
reduces no. of variables w high loadings, simplifies interpretation
what is equamax rotation?
minimises no. variables that load highly and no. factors needed to explain
What do rotations achieve?
a change of balance of the variance explained by variables but without changing the total variance
