## Some Assumptions of EFA

- Continuous variables: EFA requires the use of continuous variables, though ordinal data can be used.
- No outliers: EFA is sensitive to outliers. Outliers can distort the factor loading and loading structure.
- No multicollinearity: There should be no perfect multicollinearity among the observed variables. Multicollinearity occurs when one observed variable can be predicted from the linear combination of other observed variables.
- Unidimensionality: In EFA, each observed variable must be associated with one factor only. This variable must have a higher factor loading than the cross-loading.
- Sample size: This is very debatable. I recommend a larger sample size if possible.
- Normality: EFA does not assume the normality of the observed variable
- Linearity: EFA assumes that the linear relationship between observed variables is linear. Pearson correlations can be used to examine linear relationships.