My data set include few variables which are ratio type such as as age, income etc. Other are gender, possessing a particular asset or not. I am keen to do PCA with an objective of feature selection or feature extraction.
you can represent the nominal or the interval scale with digital number as starta or classes, such as the gender (male = 1 female = 0), scale values (class 1,2,3 ... etc.)
Dear Mohamed Mortada El- Sharkawy
Thank you very much for your answer. This conversion will not make it a continuous variable.
Regards
yes the gender data is already not continuous data, for the interval scale it will be converted to classes.
Mixing up of different level of scales (two-point) with likert scale (5-point) will lead to wrong results in PCA. PCA should be done with similar item scales and then other techniques like ANOVA, MANOVA can be utilised for further analysis.
Feature that contain interval scale can be transferred into digital like Mohamed suggested without worry. Transforming features that describe category for example gender into digits should be done very carefully because if male is set to 1 and female is set to 0 it may suggest that male is grater than female.
- Badges
- Science topic