statistic test statistics.
 
    DESCRIPTION
    Computes test statistic stat (or multiple using a cell array) for a
    nsamples x nvars design matrix D and a nsamples x nvars prediction
    matrix P.
 
    Available statistics:
 
    'accuracy'    : proportion of correctly classified samples
    'logprob'     : log probability of the correct class
    'correlation' : correlation between input and output matrices
    'contingency' : contingency matrix (rows=true class, columns=predicted class)
    'confusion'   : confusion matrix (rows=true class, columns=predicted class)
    'binomial'    : binomial test gives p value for significant
                    classification; takes uneven class distributions into
                    account
    'MAD'         : mean absolute deviation in degrees for angles specified
                    in radians
    'RMS'         : root-mean-square error
    'tlin'        : t-linear statistic to measure goodness of fit for
                    circular data with angles specified in radians 
                    using fisher's correlation coefficient.
    'R2'          : squared correlation; explained variance in the linear regression case
    'expvar'      : explicit computation of explained variance: 
                    (var(y) - var(y-yhat)) / var(y)
                    1 = perfect prediction, <0 = worse than chance
    'identity'    : computes the proportion of trials that are identified correctly 
 
    NOTE: notation '-x' with x one of the above is also allowed. This way
    error measures such as 'MAD' can be used as a performance measure by
    dml.permutation or dml.gridsearch using '-MAD'
 
    EXAMPLE
    s = dml.statistic('accuracy',D,P)
    s = dml.statistic('-MAD',D,P)
 
    DEVELOPER
    Marcel van Gerven (m.vangerven@donders.ru.nl)