Using principal component analysis to extract mixed variables for smoothed location model
This study is conducted to test the appropriateness of variables extraction technique called principal component analysis to keep adequate number of variables for construction of the smoothed location model when the measured variables are mixed and large, particularly the binary.The strategy of perf...
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Pushpa Publishing House
2013
|
| Subjects: | |
| Online Access: | https://repo.uum.edu.my/id/eprint/21572/1/FJMS%20%2080%201%202013%2033%2054.pdf |
| Summary: | This study is conducted to test the appropriateness of variables extraction technique called principal component analysis to keep adequate number of variables for construction of the smoothed location model when the measured variables are mixed and large, particularly the binary.The strategy of performing variables extraction prior to construction of the smoothed location model was tested on some artificial data generated from normal population and also on three real data sets.The criterion for selecting useful components based on unity eigenvalue works well for small and moderate sizes of variables, but troublesome for large number of variables.Results of simulations give evidence that the proposed strategy shows good performance with small value of error rate in classifying future objects through new extracted components for the investigated population.The results also are satisfactory for real data sets.The suggested strategy shows potential to be used as an option for the purpose of classification based on smoothed location model when dealing with large number of mixed variables. |
|---|