Factor Analysis Flashcards
What is the process of ‘factor loading’?
When initial predictors load on to the underlying factors. I.e. more loading happens when more initial predictors load on to an identified underlying factor
What are the two types of factor analysis based on approach?
Confirmatory
Exploratory
What is the correlation matrix and what function does it have?
It is a table showing the correlations that exist between the different predictors as well as the associated level of significance assigned to these correlation values. It serves to inform the user whether or not there is sufficient reason to conduct a factor analysis
When would it be appropriate to conduct a factor analysis based off the correlation matrix?
There are multiple cases of strong correlations between predictors that are also statistically significant
What are the two types of variance in terms of how different predictors explain the overall variance of Y? What do they mean?
Common and unique variance. Common means the proportion of explained variance in Y that is shared by both predictors i.e. they both contribute in a similar way to explaining y variance. Unique is solely held by one predictor
What type of factor analysis is principal component analysis?
Exploratory
What is the process of PCA?
It calculates the common and unique variance from scratch then constructs one table showing the different underlying factors for the predictors based off this information. It then also constructs another table to describe the factor loading process
What does the first table that SPSS produces containing the underlying factors outline?
The associated % of the variance in Y that is explained by that underlying factor.
What does the second table that SPSS produces contain?
It outlines how strongly the different predictors load on to the underlying factors i.e. which underlying factor does that predictor load on to most
Explain the ‘kaiser criterion’?
This is the criteria that PCA employs for defining where an underlying factor can be identified. Underlying factors are identified if their eigenvalue >1
What is a disadvantage of employing the kaiser criterion for identifying underlying factors?
Some underlying factors may have an eigenvalue of 0.99 which means they still explain a large amount of the variance and so if we discard them as not an underlying factor then we miss out on a lot of information that it would provide
What is the scree test?
A scree plot outlines the degree of variance (y-axis) that different identified underlying factors (x-axis) explain. Where the line reaches its inflexion point is regarded as the cut off point for identifying the underlying factors
What is a component matrix?
It is a table (the second table produced from an SPSS PCA) that outlines how strongly the different initial predictors load on to the underlying factors
What is the process of rotation?
How well different predictors load on to underlying factors can be represented on a graph. If a predictor does not load on to one underlying factor particularly strongly then we can rotate the axis via two methods to make sure that the predictor loads more strongly on to one of the underlying factors and less on to the others.
What are the two forms of rotation?
Orthogonal and oblique
