Factor analysis Flashcards
What is factor analysis?
A method where the variations in the scores are expressed in the smallest amount of data possible. It creates a set of variables for similar items that are called dimensions to
What are dimensions in factor analysis?
When factor analysis creates a set of variables for similar items in a set. The set is called a dimension
What does factor analysis try to determine?
Whether the variation in the variables can be explained by a smaller number of variables we can’t see
What are the assumptions of factor analysis? (Hunt: there are 5)
1) no outliers
2) adequate sample size
3) minimal multi collinearity
4) interval data
5) linearity of data
When would you use the exploratory factor analysis?
When you do not know how many dimensions are in your set of variables; used when you don’t know what your answer will be
What does Confirmatory factor analysis allow us to do?
It allows us to find out if there is a relationship between our set of variables and to see r there is any underlying constructs
What does multivariate correlational statistics do?
Correlates everything; it correlates one participants scores with all other participants scores
What can the variance in each question in factor analysis be explained by?
It can be explained by a common factor
What is the Beck Depression inventory?
Where you read a group of statements that belong to a question and you pick the question that most relates to you. When you get a high score this can contribute to your overall depression mark
What is a potential issue with the Beck Depression inventory?
Some questions such as “I’m more tired than usual” may be due to working long hours and not depression
What is explanatory factor analysis? What does it reduce?
It reduces data into smaller sets of summary variables.
How does factor analysis removes redundacy of variables?
It does this by removing duplication variables; it represents correlated variables with a set of variables called factors. These factors are relatively seperate from one another. They are independent
How do you check to correlations in factor analysis?
By looking at the correlation matrix or the determinant of the R-matrix
Why is Singularity of variables a problem?
When the variables are perfectly correlated it is impossible to determine which variable had what effect on the criterion variable
In SPSS, what is a useful way to establish how many factors should be kept in your factor analysis?
Looking at the scree plots.
What does rotation do in factor analysis?
Rotation increases the loading of each variable on the extracted factors while decreasing the loading of all other factors
When would you use a orthogonal rotation in factor analysis?
We would use it if we expect our factors to be independent
What type of rotation do we use for factor analysis if we expect our factors to not be independent?
We would use oblimin
What does it mean if a variable has high communality in factor analysis?
It means the variable shares more in common with the rest of the variables
How can you tell how much variability in the criterion variable is explained by the model in factor analysis? (Using spss)
You will look at the model summary table and look at the R square and adjusted r square
How to work out the critical value of sample size in factor analysis, what is the equation?
50+8x. (X is the number of predictors you had)
E.g. If you had 4 predictors then the equation is
50+8x4= 36+50= 86
When deciding whether to use R2 or adjusted r2 to see what value explains variability in criterion value, what helps us decide that?
If the sample size is GREATER or EQUAL to the critical value then you use r2
If it is LESS than the critical value then you use the adjusted r2
Who introduced factor analysis? How did he define it
Thurstone: to minimise an excess of individual measure into fewer and more meaningful dimensions
Why does factor analysis identify correlations between predictor variables?
It does this to make them into one Factor that drives their values
How do you check if variables are linear to each other in factor analysis?
Looking at scatter plots, check if they are significant
What two methods are used to see if there is multicolinarity? What value should it be to indicate multiconlineraity or signualirty?
Method 1: correlation matrix. Correlations above 0.8 show milticolinesrity/singularity
Method 2: looking at the determinant. The value should be more than 0.00001 for there to be no multicolinesrtiy
What should you consider if there is lots of variables with correlations below 0.3?
Removing them
What test can you use to determine sufficient correlations? What value I dictates that the variables are correlated to some degree?
Using Barrett’s test of sphericity. The value needs to be less than 0.5
What do we look at to check for an adequate sample size?
We look at the kaiser myer olkin. It needs to be more than 0.5 for factor analysis to be suitable
What is Eigenvalues? What is it used for and what does it indicate?
Used to condense/minimisale the variance in the correlation matrix. It indicates the amount of variance in the correlation matrix that is produced by a factor
What is the minimum eigenvalue used for extracting factors? What is it known as?
The minimum value is 1. It is called the Kaiser-Guttman rule
Why might you use PCA?
1) if you think you have too many variables and you think some are measuring the same underlying concept
You
What table do you check in SPSS output to see if the data is suitable for reduction?
Barrett’s test of spbericity