The Anova 1, Regression 1, Non-parametric 1, Distribution and Goodness of Fit examples workbooks have also been revised for compatibility with the new release. These are available for free download at Download Examples Workbooks.
Over the course of the next several days, the website will be updated for compatibility with the new release.
If you are getting value from the Real Statistics website or software, I would appreciate your donations to help offset the costs of the website by going to Please Donate.
I have been asked many times by users to add You Tube videos explaining how to use various Real Statistics capabilities. I plan to do this in 2020. I also encourage you to add videos to You Tube explaining how you have used Real Statistics, in English but also in your language.
The following is an overview of the new features in Release 6.8.
Support for new distributions
Support for the Gumbel, Logistic, Laplace, Inverse Chi-square and Inverse Gamma distributions have been added. This entails the addition of the following functions which compute the pdf, cdf and inverse of these functions:
GUMBEL_DIST, LOGISTIC_DIST, LAPLACE_DIST, ICHISQ_DIST, IGAMMA_DIST
GUMBEL_INV, LOGISTIC_INV, LAPLACE_INV, ICHISQ_INV, IGAMMA_INV
In addition, we have added the following new functions that compute the parameters for the Gumbel, Logistic, Laplace, Lognormal and Geometric distributions based on the method of moments and MLE approach:
GEOM_FIT, GUMBEL_FIT, LAPLACE_FIT, LOGISTIC_FIT, LOGNORM_FIT
GUMBEL_FITM, LAPLACE_FITM, LOGISTIC_FITM
The Distribution Fitting data analysis tool has also been enhanced to support the Gumbel, Logistic, Laplace, Lognormal and Geometric distributions.
The MEAN_DIST and VAR_DIST functions have also been updated to support these new distributions. These functions now support the following distributions: normal, lognormal, t, F, chi-square, beta, gamma, Gumbel, Laplace, Logistic, geometric, uniform, exponential, binomial, negative binomial, hypergeometric, Poisson, PERT, triangular, Skellam, inverse chi-square and inverse gamma.
Confidence Intervals for Fit Parameters
The following functions have been added that calculate the standard error and confidence intervals for parameters fitted by the method of moments or MLE approaches (as well as the regression approach for the Weibull distribution):
BETA_CONF, EXPON_CONF, GAMMA_CONF, GEOM_CONF, GUMBEL_CONF, LAPLACE_CONF, LOGISTIC_CONF, LOGNORM_CONF, NORMAL_DIST, UNIFORM_CONF, WEIBULL_CONF
Two-sample Anderson Test
Support for the two-sample Anderson-Darling test has been added via the following new functions:
AD2TEST: returns the AD statistic, p-value and critical value for the two-sample AD test
AD2PROB: returns the estimated p-value for the two-sample AD test
AD2CRIT: returns the estimated critical value for the two-sample AD test
AD2CRITX: returns the critical value for the two-sample AD test for small samples based on a table of critical values
This test is also supported on the Goodness of Fit data analysis tool.
One-sample Anderson-Darling Test Enhancements
Support for the Gumbel, Logistic and Lognormal distributions have been added to the one-sample AD test. This is reflected in the ANDERSON, ADTEST, ADPROB and ADCRIT functions.
New Anderson-Darling Test Functions for Large Samples
Previously the AD test critical values and p-values for large samples were based on a table lookup. This has now been replaced by more accurate values based on two new functions:
AD_DIST(AD) = p-value for the Anderson-Darling statistic AD
AD_INV(p) = Anderson-Darling statistic corresponding to the p-value p
These functions are now used for both the one-sample and two-sample Anderson-Darling tests.
Statistical Power for Nonparametric Tests
The MW_POWER and SR_POWER functions have been added that determine the power of the Mann-Whitney and Signed-Ranks tests. These functions use simulation to estimate the power of these tests.
Samples can be taken from the following distributions: normal, lognormal, logistic, Laplace and Gumbel. These functions can also be used to determine the minimum sample size required for these tests.
Statistical Power for Tukey HSD Post-hoc Test
The TUKEY_POWER function has been added that determines the power of the Tukey HSD post hoc test following a significant ANOVA test. These functions use simulation to estimate the power of this test.
This function can also be used to determine the sample size requirements for this test. Note that often the sample size required for the ANOVA test is less than that required for the Tukey HSD post-hoc test.
Note too that a new TUKEY(R1) function has been added that performs the Tukey HSD test for the data in R1. The output takes the form of a three-column array, whose first row is for the ANOVA test and the others are for the pairwise comparisons; the first two columns specify the indices of the pairs and the third column shows the p-value for that pair.
Passing-Bablok Regression
A new Passing-Bablock Regression data analysis tool has now been added. In addition, the following new functions have been added:
PBRegCoeff: returns the slope and intercept coefficients for Passing-Bablok regression along with the endpoints of the 1-alpha confidence intervals for these coefficients
PBTEST: returns the results of the linearity test for PB regression: h-stat, h-crit and p-value
Passing-Bablok regression is a non-parametric technique for comparing two methods (especially two measurement techniques) to see whether or not they yield similar results. This is also the motivation for Deming regression and Bland-Altman, which are already supported by Real Statistics.
Bug Fixes
In the GAMMA_FITM function when the pure argument was set to TRUE, the software acted as if it were set to FALSE, and vice versa. This has been corrected.
The ICHISQ_DIST function returned 1–p where p is the correct value. This has now been corrected.