Real Statistics Release 9.6

In the next day or two the XRealStats.xlam and Mac versions of the Rel 9.6 software will be released. Over the course of the next few days, the website will be updated for compatibility with the new release. As the website is updated the links provided below will also be updated.

I want to thank everyone who made suggestions or identified errors in the website or software. Your help has improved the utility and accuracy of Real Statistics.

I also want to express my appreciation to all of you who have donated to Real Statistics. These donations help offset the costs of maintaining the website. If you are getting value from the Real Statistic website or software, I would appreciate your donation by going to Please Donate.

The following is an overview of the new features in Release 9.6.

Negative Binomial Regression

A new Negative Binomial Regression data analysis tool has been added. There are three options for how the regression coefficients are calculated: using Newton’s method, using Solver, or by using Newton’s method where the alpha dispersion value is specified.

The following new worksheet functions are also provided: NegBinomCoeff, NegBinomCoeff0, NegBinomLL, NegBinomLL0, NegBinomLL0X, NegBinomLL1, NegBinomCov, NegBinomGrad, NegBinomParam, and NegBinomProb.

Click here for more details, although a description of the new capabilities is under development.

Poisson Regression

New PoissonLL, PoissonLL0 and PoissonLL1 functions have been added that output LL_fit, LL_min and LL_max. The new release also provides the Poisson_STest and Poisson_LMTest functions which implement the Score and Lagrange Multiplier overdispersion tests.

A new PoissonCovX function has also been added that calculates the coefficient covariance matrix using the coefficient array. Finally a new PoissonProb function has been added to determine the probability that a specified count value occurs.

This release also fixes a bug that occurs when one or more of the y values is zero. Previously the deviation values produced errors.

Also previously the AIC value reported by the Poisson Regression data analysis tool was based on the G² statistic. In this release, the AIC value is based on the LL value. In addition the BIC statistic has been added to the output.

Click here for details about Poisson regression, although the webpages corresponding to these enhancements is still under development.

Click here for details about the Poisson regression overdispersion tests.

Zero-truncated Poisson (ZTP) Regression

In this release a Zero-Truncated Poisson (ZTP) regression option has been added to the Poisson Regression data analysis tool.

To support this new capability, the following new worksheet functions have also been added:

ZTPLL, ZTPLL0, ZTPLL1, ZTPParam, ZTPCov, ZTPProb

Click here for details about Poisson regression.  This part of the website is being updated to reflect the new ZTP regression capabilities.

Zero-inflated Poisson Regression

In this release a Zero-Inflated Poisson (ZIP) regression data analysis tool has been added.

To support this new capability, the following new worksheet functions have also been added:

ZIPLL, ZIPParam, ZIPCov, ZIPPredC, ZIPPredCC, ZIPProb

In addition, a new VuongTest worksheet function has been added to implement the Vuong test to determine whether there is a significant difference between the Poisson and ZIP regression models.

Click here for details about Poisson regression.  This part of the website is being updated to reflect the new ZIP regression capabilities.

Fitting a Three-parameter Weibull Distribution

This new release adds the WEIBULL3_FIT and WEIBULL3_FITM worksheet functions that fit data to a 3-parameter Weibull distribution using the MLE and MoM approaches. Click here for more details. 

Differential Equations

In Release 9.5.5, we added support for differential equations. In this release we provide the following additional support:

First-order differential equations: We have enhanced the worksheet function DiffEq by adding a “rk4” option for the Runge-Kutta order 4 method. Click here for more details. 

Simultaneous differential equations: We now provide the worksheet function DiffEqs that outputs the solution to two simultaneous differential equations. Click here for more details. 

Second-order differential equations: We now provide the worksheet function DiffEq2 that outputs the solution to a second-order differential equation. Click here for more details. 

This release also adds a Solving Differential Equations data analysis tool. Click here for more details. 

Non-parametric Correlation Enhancements

The SCorrelTest function has been added to provide a way of testing whether Spearman’s rho is significantly different from any specified value based on a Fisher transformation. Click here for more details. 

Similarly, this release adds the KCorrelTest function to provide a way of testing whether Kendall’s tau is significantly different from any specified value based on a Fisher transformation. Click here for more details. 

Minor Enhancements

This release adds the EvalSymbolic worksheet function. E.g. if cell N6 contains the lambda formula =LN(N3)+2*N4^3-7+ABS(N3)*SQRT(N5), then the formula =EvalSymbolic(N6) would output the text =LN(x)+2*y^3-7+ABS(x)*SQRT(z).

This function is used in the Solving Differential Equations data analysis tool, described above. Click here for more details. 

This release also adds the FFTEXT worksheet function. =FFTEXT(s) is equivalent to =FTEXT(INDIRECT(s)). Click here for more details. 

This release also adds the TPOISSON_FIT worksheet function to estimate the lambda parameter of a zero-truncated Poisson distribution that best fits some data.

Bug Fixes

Fixes an error in the p-value for CATEST worksheet function for the Cochran-Armitage test. Click here for more details.

Corrects the order in which the eigenvalues appear in the D matrix when using the Spectral Decomposition option of the Matrix Operations data analysis tool. Click here for details.