factor analysis Flashcards
factor analysis is an analysis of ___
interdependence
what is analysis of interdependence
There is no dependent variable in an analysis of interdependence
what is factor analysis?
Is a method of sorting a large amount of variables into FACTORS. Clumping together a large amount of variables into factors.
what is a factor?
A factor is a super-variable that contains variables that have high intercorrelations between each other but low correlations with other groups
what is the ‘loading’ factor
The relative connection of each of the original variables to a factor is called the variable’s factoring loading on that factor.
what is the purpose of factor analysis? (3 points)
To assess the degree to which items are ‘tapping’ the same concept (or different ones)
If we have a large number of variables, factor analysis can determine the degree to which they can be reduced to a smaller set.
help make sense of complex phenomena by reducing them to a more limited number of factors
factor analysis can also help to reduce ___
multicollinearity
what are the 3 main factor analysis steps
- Correlation matrix
- Extraction of initial orthogonal factors
- Rotation to final factors
2 methods for extraction of initial orthogonal factors
principal components analysis (PCA)
principal axis factoring (PAF)
what is PCA
exploratory
what does exploratory factor analysis mean
means you have very little knowledge of the variables in question, assume there is only common variance and nothing unique
what is PAF
confirmatory - realistic assumptions from theory
what does confirmatory mean
know more about the variables, use the analysis to confirm hypotheses
what is variance
how much much a variable loads onto a particular factor
what are the 3 types of variance
common variance, specific variance (unique to that variable), error variance (fluctuations from measuring something)
what type of variance does PCA assume
common (assumes that all circles overlap)
what variance does PAF look at
common and unique (what they dont have in common)
when should you use PCA (eg what would the data have to be to use it?)
have no big assumptions in data, high correlations, large number of variables, just want to reduce the data set to smaller variables
in principal component analysis what is the number in the commonalities table and why
initial will be 1 because all variance is held in common
what is the extraction score in PAF
extraction score indicates the degree of common variance that is attributed to each variable once the analysis is complete.
do you want a high extraction score in PAF
YES the higher the better - means it is more significant
what is an eigenvalue
the eigenvalue is how much variance that particular variable accounts for. the bigger value is more significant
in SPSS any eigenvalue less than __ gets cut out of the table
1
what are the 2 processes for deciding which factors to retain?
kaisers criterion
scree test
what is kaisers criterion
if the eigenvalue is more than 1 then the value should be retained
what is the scree test
a graph plot of the component number and eigenvalue, look for the flattening of the slope (POINT OF INFLEXION) which is where you should stop adding factors.
the higher the number in the component matrix the more it ___ onto that factor
loads
why do you use rotation as a step
if there are 2 numbers close together on the component matrix then you rotate to make the distinction clearer
orthogonal rotation rotates the data at __ degrees
90 degrees
how does oblique rotation work
allows the factors to correlate (ie lets x and y axes to assume a different angle than 90 degrees)
in principle axis factoring what will the communalities table show
that all the variables do not have a value of 1 (like PCA). the extraction column indicates the degree of common variance after analysis is complete - will be lower
