import numpy as np
from scikits.learn.lda import LDA


str_data = """\
affected    proband sex a   b   c   d   e   f
F   F   M   0.450   0.410   0.420   0.400   0.440   0.510
F   F   F   0.610   0.520   0.480   0.690   0.530   0.430
T   T   F   0.570   0.430   0.320   0.620   0.420   0.480
T   F   F   0.660   0.390   0.620   0.550   0.650   0.530
T   T   M   0.550   0.490   0.640   0.630   0.640   0.670
T   F   F   0.650   0.570   0.450   0.610   0.480   0.610
T   T   M   0.550   0.560   0.490   0.540   0.510   0.570
F   F   F   0.590   0.560   0.790   0.740   0.730   0.660
F   F   M   0.700   0.580   0.610   0.820   0.720   0.690"""

all_data = [x.split() for x in str_data.split("\n")]

X = np.array([map(float, row[3:]) for row in all_data[1:]])

labels = ["/".join(row[:3]) for row in all_data[1:]]
y_lookup = dict((l, i) for i, l in enumerate(set(labels)))
y = np.array([y_lookup[l] for l in labels])




lda = LDA(n_components=2)
lda.fit(X, y)

# predict ok
y_predict = lda.predict(X)
print sum(y_predict == y) / float(y.size)

# transform has bad dimensions (not aligned for np.dot)
# but works with n_components=None
X_r = lda.transform(X)