File: manova.m

package info (click to toggle)
octave 2.0.16-2
  • links: PTS
  • area: main
  • in suites: potato
  • size: 26,276 kB
  • ctags: 16,450
  • sloc: cpp: 67,548; fortran: 41,514; ansic: 26,682; sh: 7,361; makefile: 4,077; lex: 2,008; yacc: 1,849; lisp: 1,702; perl: 1,676; exp: 123
file content (154 lines) | stat: -rw-r--r-- 4,553 bytes parent folder | download | duplicates (3)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
## Copyright (C) 1996, 1997 Kurt Hornik
##
## This program is free software; you can redistribute it and/or modify
## it under the terms of the GNU General Public License as published by
## the Free Software Foundation; either version 2, or (at your option)
## any later version.
##
## This program is distributed in the hope that it will be useful, but
## WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
## General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with this file.  If not, write to the Free Software Foundation,
## 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.

## usage:  manova (Y, g)
##
## Performs a one-way multivariate analysis of variance (MANOVA). The
## goal is to test whether the p-dimensional population means of data
## taken from k different groups are all equal.  All data are assumed
## drawn independently from p-dimensional normal distributions with the
## same covariance matrix.
##
## Y is the data matrix.  As usual, rows are observations and columns
## are variables.  g is the vector of corresponding group labels (e.g.,
## numbers from 1 to k), so that necessarily, length (g) must be the
## same as rows (Y).
##
## The LR test statistic (Wilks' Lambda) and approximate p-values are
## computed and displayed.

## Three test statistics (Wilks, Hotelling-Lawley, and Pillai-Bartlett)
## and corresponding approximate p-values are calculated and displayed.
## (Currently NOT because the f_cdf respectively betai code is too bad.)
  
## Author:  TF <Thomas.Fuereder@ci.tuwien.ac.at>
## Adapted-By:  KH <Kurt.Hornik@ci.tuwien.ac.at>
## Description:  One-way multivariate analysis of variance (MANOVA)

function manova (Y, g)

  if (nargin != 2)
    usage ("manova (Y, g)");
  endif

  if (is_vector (Y))
    error ("manova:  Y must not be a vector");
  endif

  [n, p] = size (Y);

  if (!is_vector (g) || (length (g) != n))
    error ("manova:  g must be a vector of length rows (Y)");
  endif

  s = sort (g);
  i = find (s (2:n) > s(1:(n-1)));
  k = length (i) + 1;
    
  if (k == 1)
    error ("manova:  there should be at least 2 groups");
  else
    group_label = s ([1, (reshape (i, 1, k - 1) + 1)]);
  endif

  Y = Y - ones (n, 1) * mean (Y);
  SST = Y' * Y;

  s = zeros (1, p);
  SSB = zeros (p, p);
  for i = 1 : k;
    v = Y (find (g == group_label (i)), :);
    s = sum (v);
    SSB = SSB + s' * s / rows (v);
  endfor
  n_b = k - 1;
    
  SSW = SST - SSB;
  n_w = n - k;

  l = real (eig (SSB / SSW));
  l (l < eps) = 0;

  ## Wilks' Lambda
  ## =============

  Lambda = prod (1 ./ (1 + l));
  
  delta = n_w + n_b - (p + n_b + 1) / 2
  df_num = p * n_b
  W_pval_1 = 1 - chisquare_cdf (- delta * log (Lambda), df_num);
  
  if (p < 3)
    eta = p;
  else
    eta = sqrt ((p^2 * n_b^2 - 4) / (p^2 + n_b^2 - 5))
  endif

  df_den = delta * eta - df_num / 2 + 1
  
  WT = exp (- log (Lambda) / eta) - 1
  W_pval_2 = 1 - f_cdf (WT * df_den / df_num, df_num, df_den);

  if (0)

    ## Hotelling-Lawley Test
    ## =====================
  
    HL = sum (l);
  
    theta = min (p, n_b);
    u = (abs (p - n_b) - 1) / 2; 
    v = (n_w - p - 1) / 2;

    df_num = theta * (2 * u + theta + 1);
    df_den = 2 * (theta * v + 1);

    HL_pval = 1 - f_cdf (HL * df_den / df_num, df_num, df_den);

    ## Pillai-Bartlett
    ## ===============
  
    PB = sum (l ./ (1 + l));

    df_den = theta * (2 * v + theta + 1);
    PB_pval = 1 - f_cdf (PB * df_den / df_num, df_num, df_den);

    printf ("\n");
    printf ("One-way MANOVA Table:\n");
    printf ("\n"); 
    printf ("Test             Test Statistic      Approximate p\n");
    printf ("**************************************************\n");
    printf ("Wilks            %10.4f           %10.9f \n", Lambda, W_pval_1);
    printf ("                                      %10.9f \n", W_pval_2);
    printf ("Hotelling-Lawley %10.4f           %10.9f \n", HL, HL_pval);
    printf ("Pillai-Bartlett  %10.4f           %10.9f \n", PB, PB_pval);
    printf ("\n");

  endif

  printf ("\n");
  printf ("MANOVA Results:\n");
  printf ("\n");
  printf ("# of groups:     %d\n", k);  
  printf ("# of samples:    %d\n", n);
  printf ("# of variables:  %d\n", p);
  printf ("\n");  
  printf ("Wilks' Lambda:   %5.4f\n", Lambda);
  printf ("Approximate p:   %10.9f (chisquare approximation)\n", W_pval_1);
  printf ("                 %10.9f (F approximation)\n", W_pval_2);
  printf ("\n");
  
endfunction