# [R] Ellipse Cosrrespondence Analysis(Internal Stability)

From: Aam Sudrajat <azoery011388_at_yahoo.com>
Date: Thu 14 Jul 2005 - 19:45:38 EST

I'am a student of Educational Universiti of Indonesia. I take "Internal Stabiliti in Correespondence Analysis of Normal Distributerd Data" for my final task.

I've had problem with ellipse contructing as STATISTICA 6.0 generate in Correspondence Analysis Session. I thing the ellipse in bivariate normal distribution can be applied in the plot of correspondence analysis.

The ellipse is
t(X-Xbar)(inverse(Sigma))(X-Xbar)<= chi-square(alfa,2) and the probability is
P[t(X-Xbar)(inverse(Sigma))(X-Xbar)<= chi-square(alfa,2)]=1-alfa

The point outside the ellipse become supplementary points as the point outside the ellipse in the bivariate normal is outlier.

If i construct the ellipse from the row and column coordinates like t(F-Fbar)(inverse(Sigma(F)))(F-Fbar)<= chi-square(alfa,2),

wheter the ellipse valid? (for detecting outlier data)

How can i create the valid ellipse ?