fitdistrplus (version 1.2-1)

gofstat: Goodness-of-fit statistics

Description

Computes goodness-of-fit statistics for parametric distributions fitted to a same censored or non-censored data set.

Usage

gofstat(f, chisqbreaks, meancount, discrete, fitnames=NULL) 
	
# S3 method for gofstat.fitdist
print(x, ...)
# S3 method for gofstat.fitdistcens
print(x, ...)

Value

gofstat() returns an object of class "gofstat.fitdist" or "gofstat.fitdistcens" with following components or a sublist of them (only aic, bic and nbfit for censored data) ,

chisq

a named vector with the Chi-squared statistics or NULL if not computed

chisqbreaks

common breaks used to define cells in the Chi-squared statistic

chisqpvalue

a named vector with the p-values of the Chi-squared statistic or NULL if not computed

chisqdf

a named vector with the degrees of freedom of the Chi-squared distribution or NULL if not computed

chisqtable

a table with observed and theoretical counts used for the Chi-squared calculations

cvm

a named vector of the Cramer-von Mises statistics or "not computed" if not computed

cvmtest

a named vector of the decisions of the Cramer-von Mises test or "not computed" if not computed

ad

a named vector with the Anderson-Darling statistics or "not computed" if not computed

adtest

a named vector with the decisions of the Anderson-Darling test or "not computed" if not computed

ks

a named vector with the Kolmogorov-Smirnov statistic or "not computed" if not computed

kstest

a named vector with the decisions of the Kolmogorov-Smirnov test or "not computed" if not computed

aic

a named vector with the values of the Akaike's Information Criterion.

bic

a named vector with the values of the Bayesian Information Criterion.

discrete

the input argument or the automatic definition by the function from the first object of class "fitdist" of the list in input.

nbfit

Number of fits in argument.

Arguments

f

An object of class "fitdist" (or "fitdistcens" ), output of the function fitdist() (resp. "fitdist()"), or a list of "fitdist" objects, or a list of "fitdistcens" objects.

chisqbreaks

Only usable for non censored data, a numeric vector defining the breaks of the cells used to compute the chi-squared statistic. If omitted, these breaks are automatically computed from the data in order to reach roughly the same number of observations per cell, roughly equal to the argument meancount, or sligthly more if there are some ties.

meancount

Only usable for non censored data, the mean number of observations per cell expected for the definition of the breaks of the cells used to compute the chi-squared statistic. This argument will not be taken into account if the breaks are directly defined in the argument chisqbreaks. If chisqbreaks and meancount are both omitted, meancount is fixed in order to obtain roughly \((4n)^{2/5}\) cells with \(n\) the length of the dataset.

discrete

If TRUE, only the Chi-squared statistic and information criteria are computed. If missing, discrete is passed from the first object of class "fitdist" of the list f. For censored data this argument is ignored, as censored data are considered continuous.

fitnames

A vector defining the names of the fits.

x

An object of class "gofstat.fitdist" or "gofstat.fitdistcens".

...

Further arguments to be passed to generic functions.

Author

Marie-Laure Delignette-Muller and Christophe Dutang.

Details

For any type of data (censored or not), information criteria are calculated. For non censored data, added Goodness-of-fit statistics are computed as described below.

The Chi-squared statistic is computed using cells defined by the argument chisqbreaks or cells automatically defined from data, in order to reach roughly the same number of observations per cell, roughly equal to the argument meancount, or sligthly more if there are some ties. The choice to define cells from the empirical distribution (data), and not from the theoretical distribution, was done to enable the comparison of Chi-squared values obtained with different distributions fitted on a same data set. If chisqbreaks and meancount are both omitted, meancount is fixed in order to obtain roughly \((4n)^{2/5}\) cells, with \(n\) the length of the data set (Vose, 2000). The Chi-squared statistic is not computed if the program fails to define enough cells due to a too small dataset. When the Chi-squared statistic is computed, and if the degree of freedom (nb of cells - nb of parameters - 1) of the corresponding distribution is strictly positive, the p-value of the Chi-squared test is returned.

For continuous distributions, Kolmogorov-Smirnov, Cramer-von Mises and Anderson-Darling and statistics are also computed, as defined by Stephens (1986).

An approximate Kolmogorov-Smirnov test is performed by assuming the distribution parameters known. The critical value defined by Stephens (1986) for a completely specified distribution is used to reject or not the distribution at the significance level 0.05. Because of this approximation, the result of the test (decision of rejection of the distribution or not) is returned only for data sets with more than 30 observations. Note that this approximate test may be too conservative.

For data sets with more than 5 observations and for distributions for which the test is described by Stephens (1986) for maximum likelihood estimations ("exp", "cauchy", "gamma" and "weibull"), the Cramer-von Mises and Anderson-darling tests are performed as described by Stephens (1986). Those tests take into account the fact that the parameters are not known but estimated from the data by maximum likelihood. The result is the decision to reject or not the distribution at the significance level 0.05. Those tests are available only for maximum likelihood estimations.

Only recommended statistics are automatically printed, i.e. Cramer-von Mises, Anderson-Darling and Kolmogorov statistics for continuous distributions and Chi-squared statistics for discrete ones ( "binom", "nbinom", "geom", "hyper" and "pois" ).

Results of the tests are not printed but stored in the output of the function.

References

Cullen AC and Frey HC (1999), Probabilistic techniques in exposure assessment. Plenum Press, USA, pp. 81-155.

Stephens MA (1986), Tests based on edf statistics. In Goodness-of-fit techniques (D'Agostino RB and Stephens MA, eds), Marcel Dekker, New York, pp. 97-194.

Venables WN and Ripley BD (2002), Modern applied statistics with S. Springer, New York, pp. 435-446, tools:::Rd_expr_doi("10.1007/978-0-387-21706-2").

Vose D (2000), Risk analysis, a quantitative guide. John Wiley & Sons Ltd, Chischester, England, pp. 99-143.

Delignette-Muller ML and Dutang C (2015), fitdistrplus: An R Package for Fitting Distributions. Journal of Statistical Software, 64(4), 1-34, tools:::Rd_expr_doi("https://doi.org/10.18637/jss.v064.i04").

See Also

fitdist.

Examples

# (1) fit of two distributions to the serving size data
# by maximum likelihood estimation
# and comparison of goodness-of-fit statistics
#

data(groundbeef)
serving <- groundbeef$serving
(fitg <- fitdist(serving, "gamma"))
gofstat(fitg)
(fitln <- fitdist(serving, "lnorm"))
gofstat(fitln)

gofstat(list(fitg, fitln))


# (2) fit of two discrete distributions to toxocara data
# and comparison of goodness-of-fit statistics
#

data(toxocara)
number <- toxocara$number

fitp <- fitdist(number,"pois")
summary(fitp)
plot(fitp)

fitnb <- fitdist(number,"nbinom")
summary(fitnb)
plot(fitnb)

gofstat(list(fitp, fitnb),fitnames = c("Poisson","negbin"))

# (3) Get Chi-squared results in addition to
#     recommended statistics for continuous distributions
#

set.seed(1234)
x4 <- rweibull(n=1000,shape=2,scale=1)
# fit of the good distribution
f4 <- fitdist(x4,"weibull")
plot(f4)

# fit of a bad distribution
f4b <- fitdist(x4,"cauchy")
plot(f4b)

(g <- gofstat(list(f4,f4b),fitnames=c("Weibull", "Cauchy")))
g$chisq
g$chisqdf
g$chisqpvalue
g$chisqtable

# and by defining the breaks
(g <- gofstat(list(f4,f4b), 
chisqbreaks = seq(from = min(x4), to = max(x4), length.out = 10), fitnames=c("Weibull", "Cauchy")))
g$chisq
g$chisqdf
g$chisqpvalue
g$chisqtable

# (4) fit of two distributions on acute toxicity values 
# of fluazinam (in decimal logarithm) for
# macroinvertebrates and zooplancton
# and comparison of goodness-of-fit statistics
#

data(fluazinam)
log10EC50 <-log10(fluazinam)
(fln <- fitdistcens(log10EC50,"norm"))
plot(fln)
gofstat(fln)
(fll <- fitdistcens(log10EC50,"logis"))
plot(fll)
gofstat(fll)

gofstat(list(fll, fln), fitnames = c("loglogistic", "lognormal"))