significance

Statistical significance adjusted for multiple hypotheses

Syntax

   sig = significance(pval,alpha,correction)

Arguments

See also Common variable names and data structures.

input

   pval         multi-trial time series data
   alpha        significance level
   correction   multiple hypotheses correction (see Description)

output

   sig          significance (0 or 1)

Description

Returns significance (0 or 1) of statistics based on p-values in pval, which may be a scalar, vector or matrix. NaNs are ignored. The correction parameter specifies a multiple hypotheses test adjustment, and may be one of: 'None', 'Bonferroni', 'Sidak', 'Holm', 'FDR' (false discovery rate, independent or positively correlated hypotheses [1]) or 'FDRD' (false discovery rate, arbitrary dependencies [2]).

Note: correction = 'None' is not recommended for multiple hypotheses, so is not the default! 'FDR' is generally a good choice.

References

[1] Y. Benjamini and Y. Hochberg, "Controlling the false discovery rate: a practical and powerful approach to multiple testing", J. Royal Stat. Soc. B, 57(1), 1995.

[2] Y. Benjamini and D. Yekutieli, "The control of the false discovery rate in multiple testing under dependency", Ann. Stat, 29(4), 2001.

See also

mvgc_pval | empirical_pval