SPRING Factor Analysis

Question 1

what is the purpose of factor analysis

Answer 1

data reduction - understand the variables and their relationships and reduce to their underlying factors
aim to explain variance between individuals and certain variables- account for as much variance as possible

Question 2

what is a variable

Answer 2

score/measure of an underlying construct

Question 3

what is a factor

Answer 3

clusters of variables
a dimension/component of a construct ie introversion/extravarsion of personality
beliefs/attitudes etc
some corr more highly than others

Question 4

what is a latent variable

Answer 4

what youre expecting to measure

want to identify clusters of factros that accurately measure what you want them to

Question 5

what are pure factors/clusters of factors

Answer 5

uncorrelated - not too strongly with some and weaky with others

Question 6

how does factor analysis mean to account for as much variance as possible

Answer 6

gather small no factors that explain a good proportion of the vairance
not always able to explain all variables but can explain majoirty well

Question 7

what is a principle component analysis PCA

Answer 7

simplest factor analysis (factor = component)

Question 8

when can you use PCA

Answer 8

correlational (interval/ordinal NOT categorical)
sample size 300+
min 5 but preferaby 10+ subjects per variable
cases w/missing data omitted - need to know how many actually analysed

Question 9

why might you have missing data

Answer 9

too sensitive
missed by accident
didnt understand the question
code data dependent on diff reasons

Question 10

how many factors need to explain the majority of vairance?

Answer 10

only keep those that explain the most

eigenvalue > 1+ + work out cumulative variance

Question 11

what is cumulative variance

Answer 11

total variance that all the variables explain together

Question 12

communalities

Answer 12

proportion of factors in each variable explained by all the extracted factors/components
want variables to explain more that half (0.5) of communalities (extraction collumn in spss) BUT subjective cut off
therefore variables w/communialities 0.5+ are reasonably well explained by the extracted factors

Question 13

component analysis

Answer 13

shows the initial factors and their loadings- the correlation between the variables and the factors
a strong loading (+-) means variable is strongly representative of a factor

Question 14

loading values

Answer 14

0. 6 = strong
1. 4 = mod
2. 3 = weak

Question 15

rotating extracted factors

Answer 15

difficult to interpret
factors in a graph tend to clster together
rotate axic so that loads better on one axis - easier
shows how each variable is represented on each of the 4 factors

Question 16

rotation and pure loadings

Answer 16

each factor made to correlate purely with a small seperate set of variables via rotation

Question 17

what is a pure loading

Answer 17

6+ on one factor and

Question 18

independence of factors

Answer 18

varimax rotated factors designed to be uncorrelated
score on one factor should have little/no corr with other factor
can use for ANOVA

Question 19

PCA and demographics

Answer 19

same analysis with added variables

- will adding these impact loading on certain factors?

Question 20

exploratory factor analysis

Answer 20

how many factors are there and what variables belong to them

Question 21

confirmatory factor analysis

Answer 21

check data contains a specified set of factors