Statistics[ChiSquareSuitableModelTest]  apply the chisquare suitable model test

Calling Sequence


ChiSquareSuitableModelTest(X, F, options)


Parameters


X



observed data sample

F



function, algebraic; probability distribution or random variable to match data against

options



(optional) equation(s) of the form option=value where option is one of bins, fittedparameters, level, output, or range; specify options for the ChiSquareSuitableModelTest function





Description


•

The ChiSquareSuitableModelTest function performs the chisquare suitable model test on an observed data sample against a known random variable or probability distribution. The test attempts to determine if the given sample can be considered to have been drawn from the given random variable or probability distribution by constructing probability categories and applying a goodnessoffit test

•

The first parameter X is a data set of observed data to use in the analysis.

•

The second parameter F is a random variable or probability distribution that is compared to the observed data set. If any parameters are symbolic, then MaximumLikelihoodEstimate is used to estimate them. This is taken into account when computing the degrees of freedom; see the description of the fittedparameters option.

•

As much as possible, the bins are chosen so that the expected number of points in each bin is the same, because if there are bins where this number is very small, the test does not perform well. This is relatively straightforward if F describes a continuous random variable, but if it is a discrete random variable this is not the case.



Options



The test_options argument can contain one or more of the options shown below.

•

bins='deduce' or posint


This option indicates the number of bins to use when categorizing data from X and probabilities from F. If set to 'deduce' (default), the function attempts to determine a reasonable value for this option.


If F describes a discrete random variable, the final number of bins may not be exactly equal to the value of the option given.

•

fittedparameters=nonnegative integer

•

range='deduce' or range


This option indicates the range to use when considering data values  data outside of the range is discarded during processing. If set to 'deduce' (default), the function attempts to determine a suitable range.


This option is used to specify the level of the analysis (minimum criteria for the observed data to be considered wellfit to the expected data). By default, this value is 0.05.

•

output='report', 'statistic', 'pvalue', 'criticalvalue', 'distribution', 'hypothesis', or list('statistic', 'pvalue', 'criticalvalue', 'distribution', 'hypothesis')


This option is used to specify the desired format of the output from the function. If 'report' is specified then a module containing all output from this test is returned. If a single parameter name is specified other than 'report' then that quantity alone is returned. If a list of parameter names is specified then a list containing those quantities in the specified order will be returned.



Notes


•

This test generates a complete report of all calculations in the form of a userinfo message. In order to access this report, specify infolevel[Statistics] := 1.



Compatibility


•

The Statistics[ChiSquareSuitableModelTest] command was updated in Maple 18.

•

The F parameter was updated in Maple 18.

•

The fittedparameters option was introduced in Maple 18.



Examples


>


>


Initialize an array of data
>


Perform the suitable model test on this sample. The null hypothesis in both cases is that the data came from the specific distribution given, with the given parameter values.
>


ChiSquare Test for Suitable Probability Model

Null Hypothesis:
Sample was drawn from specified probability distribution
Alt. Hypothesis:
Sample was not drawn from specified probability distribution
Bins: 10
Degrees of freedom: 9
Distribution: ChiSquare(9)
Computed statistic: 4.4
Computed pvalue: 0.883171
Critical value: 16.9189774487099
Result: [Accepted]
There is no statistical evidence against the null hypothesis
 
 (1) 
>


ChiSquare Test for Suitable Probability Model

Null Hypothesis:
Sample was drawn from specified probability distribution
Alt. Hypothesis:
Sample was not drawn from specified probability distribution
Bins: 10
Degrees of freedom: 9
Distribution: ChiSquare(9)
Computed statistic: 169
Computed pvalue: 0
Critical value: 16.9189774487099
Result: [Rejected]
There exists statistical evidence against the null hypothesis
 
 (2) 
If we test whether the data could come from any uniform or normal distribution, we get different numbers, and the test is no longer able to exclude the possibility that the data came from a normal distribution.
>


ChiSquare Test for Suitable Probability Model

Null Hypothesis:
Sample was drawn from specified probability distribution
Alt. Hypothesis:
Sample was not drawn from specified probability distribution
Bins: 10
Degrees of freedom: 7
Distribution: ChiSquare(7)
Computed statistic: 5.6
Computed pvalue: 0.587151
Critical value: 14.0671405764057
Result: [Accepted]
There is no statistical evidence against the null hypothesis
 
 (3) 
>


ChiSquare Test for Suitable Probability Model

Null Hypothesis:
Sample was drawn from specified probability distribution
Alt. Hypothesis:
Sample was not drawn from specified probability distribution
Bins: 10
Degrees of freedom: 7
Distribution: ChiSquare(7)
Computed statistic: 11.8
Computed pvalue: 0.107331
Critical value: 14.0671405764057
Result: [Accepted]
There is no statistical evidence against the null hypothesis
 
 (4) 
If we obtain the parameters for the distribution, from the data set, we need to specify this using the fittedparameters option.
>


 (5) 
>


ChiSquare Test for Suitable Probability Model

Null Hypothesis:
Sample was drawn from specified probability distribution
Alt. Hypothesis:
Sample was not drawn from specified probability distribution
Bins: 10
Degrees of freedom: 7
Distribution: ChiSquare(7)
Computed statistic: 10.2
Computed pvalue: 0.17752
Critical value: 14.0671405764057
Result: [Accepted]
There is no statistical evidence against the null hypothesis
 
 (6) 
Note that the value is slightly different from the previous example, because the maximum likelihood estimate for the parameter is the uncorrected sample standard deviation, in contrast to the result of the command: that is the corrected sample standard deviation (the square root of the unbiased estimator for the variance).


References



Kanju, Gopal K. 100 Statistical Tests. London: SAGE Publications Ltd., 1994.


Sheskin, David J. Handbook of Parametric and Nonparametric Statistical Procedures. London: CRC Press, 1997.


