UCS::AM::Parametric - Parametric association measures
use UCS; use UCS::AM::Parametric; @parametric_AMs = UCS::AM_Keys(); # MI.conf # MI.conf.<n> [<n> = 2, 3, 5, 10, 50, 100, 1000] # Poisson.mu.pv # Poisson.mu.<n>.pv [<n> = 2, 3, 5, 10, 50, 100, 1000, 10000]
This module contains some parametric association measures, which are parametrised extensions of measures defined in the basic UCS::AM module. Parametric measures are a recent development in cooccurrence statistics, and the choice of appropriate parameter values is still very much a research question. Parametric measures will often be computationally expensive and may be numerically unstable, so novice users are advised to use the basic measures from the UCS::AM module instead.
The following section gives a full listing of the parametric association measures defined in the UCS::AM::Parametric module with short explanations. Please refer to http://www.collocations.de/AM/ for the full equations and references. When the module is imported, the additional measures are registered with the UCS core library (see the UCS manpage for details on how to access registered association measures).
The association scores of measures with the suffix .pv can be interpreted as probabilities (i.e. the likelihood of the observed data or the p-value of a statistical hypothesis test). Such probabilities are given as negative base 10 logarithms, ranging from 0 to +inf (+inf is represented by the return value of the built-in inf function (see the UCS::Expression::Func manpage).
Conservative estimate for the base 10 logarithm of the mu-value (whose maximum-likelihood estimate is given by the MI measure). The association score computed by MI.conf is the lower endpoint of a two-sided confidence interval for mu at significance level alpha, which is specified by the ALPHA parameter (as a negative base 10 logarithm). The "usual" significance levels .01 and .001 correspond to ALPHA=2 and ALPHA=3, respectively.
Please duplicate the UCS::Expression object returned by UCS::AM_Expression("MI.conf") before modifying the ALPHA parameter.
Versions of MI.conf with the ALPHA parameter pre-set to the value specified as part of the name. Available ALPHA values are 2, 3, 5, 10, 50, 100, and 1000. For instance, MI.conf.10 computes a two-sided confidence interval at significance level 1E-10.
Do not modify the ALPHA parameter of these association measures (in the UCS::Expression object returned by the UCS::AM_Expression function).
Poisson test for O11 under the modified point null hypothesis pi = p1 * p2 * mu (rather than the independence hypothesis pi = p1 * p2 used by the Poisson.pv measure). The (non-logarithmic) value of mu is given by the MU parameter. For MU=1, the association scores computed by Poisson.mu.pv are identical to those of Poisson.pv.
Please duplicate the UCS::Expression object returned by UCS::AM_Expression("Poisson.mu.pv") before modifying the MU parameter.
Versions of Poisson.mu.pv with the MU parameter pre-set to the value specified as part of the name. Available MU values are 2, 3, 5, 10, 50, 100, 1000, and 10000.
Do not modify the MU parameter of these association measures (in the UCS::Expression object returned by the UCS::AM_Expression function).
Copyright 2003 Stefan Evert.
This software is provided AS IS and the author makes no warranty as to its use and performance. You may use the software, redistribute and modify it under the same terms as Perl itself.