Covariate-adjusted matrix visualization via correlation decomposition

Han-Ming Wu¹,Yin-Jing Tien², Meng-Ru Ho^{3, 4, 5}, Hai-Gwo Hwu⁶, Wen-chang Lin⁵, Mi-Hua Tao⁵, and Chun-Houh Chen^2,*

¹Department of Mathematics, Tamkang University, Taipei County 25137, Taiwan, R.O.C.,
²Institute of Statistical Science, Academia Sinica, Taipei 11529, Taiwan, R.O.C.,
³Institute of Biomedical Informatics, National Yang-Ming University, Taipei 112, Taiwan, R.O.C.,
⁴Bioinformatics Program, Taiwan International Graduate Program, Academia Sinica, Taipei 115, Taiwan, R.O.C.,
⁵Institute of Biomedical Sciences, Academia Sinica, Taipei 115, Taiwan, R.O.C.,
⁶Department of Psychiatry, National Taiwan University Hospital and College of Medicine, Taipei 100, Taiwan, R.O.C.
and Department of Psychology, College of Public Health, Neurobiology and Cognitive Science Center, Taipei 100, Taiwan, R.O.C.

*to whom correspondence should be addressed.

The Fisher's iris data

1. WABA and z-score significant map　

(1.1) Import the built-in Fisher's iris data.

(1.2) Click the icon for covariate adjustment (WABA analysis).

(1.3) In the Decomposition of Correlation Matrices Dialog Window, specify the covariate to be adjusted.

(1.5) The the WABA Window, one can reorder the rows and columns of the specified matrix using HCT or R2E.

2. Covariate-adjusted MV　

(2.1) Click the icon for GAP analysis.　

(2.2) In the Generalized Association Plots Dialog Window, specify the data type, the proximity measures and the covariate adjustment option.　

(2.3) In the Generalized Association Plots Window, use the control panel to zoom and coloring.

(2.4) Perform the average-linkage clustering for rows and ellipse seriation for columns.

(2.5) One can do the GAP analysis for the residual data.

(2.6) One can also do the GAP analysis for the fitted data.

Last Update: 2010/12/22, hmwu@mail.tku.edu.tw