Factor Analysis Flashcards
Background
Trying to find an underlying structure to data. Look at how diff scales correlate together then group them into factors e.g. social skills , interest in others and talking about others could be grouped into a factor called social skills . Factors are orthogonal (not linked)
Elements of analysis
Need a way to extract factors: principle component analysis and factor rotation. Something to help make judgements about factors: eigenvectors and values. Something to compare items: factor loading
Principle component analysis
Pulls out patterns in the data, finds principal axes of the data, at right angles to each other
Sometimes called eigenvectors (the maths concept based on the data) each has an eigenvakue which tells you how important it is. If above 1, sig (kaisers extraction), means factor is important. Can also use scree plot to look for inflection point
Factor loading
Once have factors, also get loadings - says how much each element relates to the factors from 0-1, high loading is above 0.5 but 0.3 still sig
Each element can load onto more than one factor
Rotation and interpretation
Factor rotation makes loading more clear, maxes the loading of a variable onto a signal factor. Two ways: varimax for uncorrelated factors /orthogonal or oblimin for related factors/oblique
. Subjective as you have to allocate factor names
Checks and balances
Variables need to be correlated (0.3 or above) but not too much e.g. no multicollinearity. Bartlett’s test of sphericity should be sig, multi should be over 0.00001 using determinant. Need large sample 100 low, 300 good, min 2ps per variable. KMO kaiser Meyer olkin says how good sample is, needs to be above 0.5
Factor scores
Can get scores for each factor for each p e.g. agreeableness, openness score for each person
Can use these as variables in a new analysis to predict something e.g wellbeing
