[pymvpa] custom cross-validation procedure: train on individual blocks, test on averaged blocks?

[Yaroslav Halchenko]

oy -- would you mind doing

print dataset.summary()

otherwise my visual system is a bit confused, although catches that
conditions seems to be not equally represented across chunks

>    after the warning message,

>    the output of dataset.summary() is:

>    "Dataset: 54x130 at float64, <sa: blocks,censor,TR,chunks,TR2,targets>, <a:
>    mapper>\nstats: mean=0.0785296 std=0.720478 var=0.519089 min=-2.79133
>    max=3.00064\n\nCounts of targets in each chunk:\n  chunks\\targets  HI 
>    HO\n                 --- ---\n       0.0        3   2\n       1.0       
>    3   3\n       2.0        3   3\n       3.0        3   3\n       4.0       
>    2   3\n       5.0        3   2\n       6.0        3   3\n       7.0       
>    3   2\n       8.0        3   3\n       9.0        2   2\n\nSummary for
>    targets across chunks\n  targets mean  std min max #chunks\n    HI    
>    2.8  0.4  2   3     10\n    HO     2.6 0.49  2   3     10\n\nSummary for
>    chunks across targets\n  chunks mean std min max #targets\n    0     2.5
>    0.5  2   3      2\n    1      3   0   3   3      2\n    2      3   0   3  
>    3      2\n    3      3   0   3   3      2\n    4     2.5 0.5  2   3     
>    2\n    5     2.5 0.5  2   3      2\n    6      3   0   3   3      2\n   
>    7     2.5 0.5  2   3      2\n    8      3   0   3   3      2\n    9     
>    2   0   2   2      2\nSequence statistics for 54 entries from set ['HI',
>    'HO']\nCounter-balance table for orders up to 2:\nTargets/Order O1     | 
>    O2     |\n     HI:      14 14  |  13 15  |\n     HO:      14 11  |  15  9 
>    |\nCorrelations: min=-0.26 max=0.18 mean=-0.019 sum(abs)=6.6"

>    here's another instance (different targets):

>    "Dataset: 57x130 at float64, <sa: blocks,censor,TR,chunks,TR2,targets>, <a:
>    mapper>\nstats: mean=0.00158443 std=0.707033 var=0.499896 min=-3.96051
>    max=2.82313\n\nCounts of targets in each chunk:\n  chunks\\targets  VI 
>    VO\n                 --- ---\n       0.0        3   3\n       1.0       
>    2   2\n       2.0        3   3\n       3.0        3   3\n       4.0       
>    3   2\n       5.0        3   3\n       6.0        3   3\n       7.0       
>    3   3\n       8.0        3   3\n       9.0        3   3\n\nSummary for
>    targets across chunks\n  targets mean std min max #chunks\n    VI     2.9
>    0.3  2   3     10\n    VO     2.8 0.4  2   3     10\n\nSummary for chunks
>    across targets\n  chunks mean std min max #targets\n    0      3   0   3  
>    3      2\n    1      2   0   2   2      2\n    2      3   0   3   3     
>    2\n    3      3   0   3   3      2\n    4     2.5 0.5  2   3      2\n   
>    5      3   0   3   3      2\n    6      3   0   3   3      2\n    7     
>    3   0   3   3      2\n    8      3   0   3   3      2\n    9      3   0  
>    3   3      2\nSequence statistics for 57 entries from set ['VI',
>    'VO']\nCounter-balance table for orders up to 2:\nTargets/Order O1     | 
>    O2     |\n     VI:      12 17  |  12 16  |\n     VO:      17 10  |  17 10 
>    |\nCorrelations: min=-0.26 max=0.37 mean=-0.018 sum(abs)=7"

>    thanks!
>    -edmund

-- 
=------------------------------------------------------------------=
Keep in touch                                     www.onerussian.com
Yaroslav Halchenko                 www.ohloh.net/accounts/yarikoptic