Analysing Discrimination Discriminant Analysis is a widely used statistical technique used to measure the relationship between 2 sets of variables. One variable is the dependent variable which varies with a set of independent variables. The method is similar to ANOVA , the only difference being the nature of the dependent variable.
Discriminant Analysis is an oft used technique to predict the nature or the belonging of a variable to a group and how it varies according to other prevalent conditions. For example, me being a Computer Science Engineer, I can relate the theory to my final year project. My final year project was about the Route Reconstruction using Support Group Method for Mobile Ad-Hoc Networks. Big name, I know. The point is that the project performed a reconstruction of signal paths between a mobile and its tower if the mobile user either moves between 2 tower zones or loses his/her current signal.
To begin with, the project calculated various causes that caused signal levels to drop and then calculated threshold values between which the signal route is reconstructed between two towers. Elementary, my dear Watson! Not so much at the time, believe me !
The most common use of Discriminant Analysis is used in applications like Face Recognition, Speech Recognition, Handwriting Analysis etc. In the financial world it is applied to real world cases to analyse bank failures, loan defaults etc. More importantly, the probability of default, most commonly measured using the Altman's Z-Score Co-efficient is nothing but a product of Discriminant Analysis.
The Limitations for this model arise when there is a huge set of independent variables. In many cases, not all independent variables might affect the dependent variable much in which case only the crucial variables need to be taken. This though changes the accuracy of the model and there is no specific number to be attained for the accuracy of a model.
Maybe, if the model were that strong, the 2008 recession might not have happened ! Guess it wasn't strong enough !
Group: Finance-2
Author: Kshitij Sharma
No comments:
Post a Comment