Real Statistics Release 9.7

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 9.7.

Multivariate Regression

Adds a new Excel worksheet function, MRegCoeff, that returns the coefficients of a multivariate regression model. This release also adds functions (MRegPred and MRegPredC) to make predictions based on a multivariate regression model.

Also provides new Excel worksheet functions (MRegSig, MRegSig0, MRegSig1, MRegSigX) that perform various significance tests of the multivariate regression coefficients using Wilk’s Lambda, Hotelling-Lawley Trace, Pillai-Bartlett Trace, and Roy’s Largest Root statistics.

To support these function the following table lookup functions have also been added: WilkCRIT, HotelCRIT, PillaiCRIT, and RoyCRIT.

Click here for more details about all these functions.

Partial Least Squares (PLS) Regression

Adds new Excel worksheet function to support PLS regression. These include PLSRegCoeff that returns the PLS regression coefficients, PLSRegVar, that displays the percentage of X and Y variances attributed to each of the latent vectors, and PLSRegPRESS, that helps determine how many latent vectors to retain.

Click here for more details about all these functions.

Bootstrapping for Regression Models

Multiple Regression

Adds new Excel worksheet functions that return the coefficient covariance matrix (RegCovBoot, RegCovBootRes) as well as the confidence intervals for the regression coefficients (RegCIBoot, RegCIBootRes) using bootstrapping. Both bootstrapping using resampling cases and resampling residuals are supported.

This new release also adds functions that return the standard error of predictions as well as the confidence and prediction intervals using bootstrapping (RegPredBoot,  RegPredBootRes). 

See Bootstrapping Regression Support for more information  about these new functions.  

This release also adds the RegPredPI worksheet function which returns predictions, standard errors, and prediction intervals using a normal approximation (without bootstrapping). It also adds the RegCoeffCI worksheet function which returns the coefficients, standard errors, t-stat, p-value, and the confidence intervals using a normal approximation (without bootstrapping).

Multivariate Regression

This release also provides similar bootstrapping support for multivariate regression models. The new worksheet functions are MRegCovBoot, MRegCovBootRes, MRegCIBoot, MRegCIBootRes, MRegPredBoot, MRegPredBootRes.

See Multivariate Regression Bootstrapping Support for more information  about these new functions.

PLS Regression

This release also provides similar bootstrapping support for PLS regression models. The new worksheet functions are PLSRegCovBoot, PLSRegCovBootRes, PLSRegCIBoot, PLSRegCIBootRes, PLSRegPredBoot, PLSRegPredBootRes.

See PLS Regression Bootstrapping for more information  about these new functions.

LASSO Regression

This release adds a LASSO Regression data analysis tool to the Reg tab.

See LASSO Regression for more details.

Enhanced Negative Binomial distribution support

The standard NEGBINOM.DIST(number_f, number_s, probability_s, cumulative) function assumes that the first argument is a non-negative whole number and the second argument is a positive whole number. If not, these arguments are rounded down to the next lower integer.

This release adds a new NEGBINOM_DIST function that accepts non-integer arguments without rounding them down. NEGBINOMX_INV has also been added as the inverse of NEGBINOM_DIST.

NEGBINOM_INV remains as the inverse of NEGBINOM.DIST.

Click here for more details about these worksheet functions.

Enhanced Frequence Table tool

The Frequency Table data analysis tool has been modified by the addition of a Freq Range. Previously, you would enter a range with two columns into the Input Range. The first column would contain the data elements and the second column would contain the frequencies of these data elements. You can still do this (leaving the Freq Range empty), but now you can optionally enter one column ranges in each field. This means that the data and frequency ranges don’t have to be adjacent.

Click here for more details.

Minor Enhancments and Bug Fixes

• Supports PseudoInv(R1, iter, prec) when R1 has one column or one row
• Supports SVD(R1) when R1 has one column or one row
• Fixes the standard error for Cohen’s Kappa test in the Interrater Reliability data analysis tool.
• Fixes errors when choosing the Paired Samples option of the T Tests and Non-Parametric Equivalents for data with blank entries.
• The SCORREL worksheet function was inadvertantly omitted from the Mac version of Rel 9.6. This function has been restored in this release.