Genstat 22 – What’s new, for you

Genstat 22 – What’s new, for you

Dr. David Baird

13 July 2022
image_blog

Equivalence tests, Linear mixed models for censored data, permutation tests for fixed effects in a GLMM and greater functionality: Genstat 22 brings a wealth of new features and benefits.

Equivalence tests

Genstat 22 has made it easy to perform equivalence, non-inferiority and non-superiority tests: extremely useful tests when the aim of the study is to demonstrate that two treatments are effectively the same, or that one treatment is neither inferior or superior to another. For example, in a medical trial when the aim is to prove that a new drug treatment is just as effective as the standard drug, or in the plant breeding context when the aim is to show that a new cultivar is at least as resistant to disease as the standard industry variety. 

Why are these tests so important? In our typical statistical hypothesis testing scenario, the null hypothesis is that of no effect, or there is no difference between two treatment means. The p-value assesses the weight of evidence against the null. However, a frighteningly common mistake is that non-significant p-values are often misinterpreted as evidence that the null hypothesis is true, leading to incorrect and potentially dangerous conclusions. So how do we demonstrate two treatment means are essentially the same? Genstat’s new and easy to perform, equivalence test facilities solve this problem. Similarly, Genstat’s new non-superiority test facilities enable us to demonstrate that one treatment is not superior to another, and the new non-inferiority test facilities that one treatment is not inferior to another.

These tests are available in both the T-test and Analysis of Variance menus.

Watch the Equivalence test video

Linear mixed models with censoring

Genstat provides everything users need to design experiments, analyse data, and generate insights - all in one easy-to-use and reliable statistical software tool. One of its many strengths is its powerful set of linear mixed models tools and its efficient REML algorithm for fitting these models. New to Genstat 22, users can now analyse censored data using a linear mixed model, and obtain well-behaved and reliable estimates of the treatment means. During data collection, censoring occurs when measurements cannot be taken above or below a bound. For example, in chemical assays where small amounts cannot be detected. Observations that fall below this minimum level of detection are therefore censored. In Genstat 22, users can now easily and quickly fit linear mixed models to such censored data. 

Watch the Linear mixed models with censoring video.

Updated Generalized linear mixed model menu

Genstat 22 offers a raft of new facilities for generalized linear mixed model (GLMM) analysis, including the ability to assess the importance of the fixed effects using permutation tests. A well-known problem with GLMMs is that the estimates of the variance components are generally smaller than the true values. The Wald tests for the fixed effects also suffer from this bias in that their p-values may be too small. Therefore, caution is needed when interpreting the Wald test p-values for fixed effects. The new GLMM permutation tests in Genstat 22 help overcome this problem by providing an alternative, more reliable, way to assess the fixed effects.

The Generalized linear mixed model (GLMM) menus have been updated to allow nearly all the same options as the Generalized linear model (GLM) menu. GLMM provides the extension of the GLM model for non-normal data to allow for multi-level variation as in REML. 

Watch the Generalized linear mixed model video.

Confidence region plot

The new Confidence region plot menu allows you to draw 2-D scatter plots with confidence, prediction and/or equal-frequency ellipses superimposed. This plot lets you look at the relationship between two variables and plot the confidence regions for this.

Watch the Confidence region plot video.

Improved importation options for Excel files

Genstat has long been able to read in data from Excel using menus or the IMPORT command. However, cell colours could not be imported. The 22nd edition can now import the cell foreground and background colours for display in the Genstat spreadsheet. In addition, it can now also import cell formulae and formatting information such as colours, fonts and styles as data. This may be useful where users highlight or format cells as a way of conveying supplementary information, such as the cell being an outlier.

Watch the Improved importation options for Excel files video.

VPERMTEST

The conventional way to assess fixed terms is to use either Wald or F tests. However, the Wald test generates significant results too frequently leading to misleading conclusions, and whilst the F test is more reliable, the denominator degrees of freedom needs to be estimated, and this is not always possible. Our new procedure, VPERMTEST, provides an alternative method for assessing fixed terms.

Watch the VPERMTEST video.

Want to know more?

There are many more new features in Genstat 22 to help users achieve the most value from their data. Chosen by thousands of scientists and researchers across the globe, Genstat’s user-friendly menu interface provides all the standard analyses, guiding non-technical users through the correct and efficient use of statistics. And for more advanced users, Genstat offers a powerful programming language that can be used to perform complex analyses, implement a new or non-standard statistical technique, or automate a task.

For a complete description of all new features and enhancements please visit What’s new in Genstat 22nd edition.

If you would like to take a look into Genstat further or obtain a quote, please email us: info@vsni.co.uk

About the author (and Genstat lead developer)

Dr. David Baird is an experienced biometrician with 40 years’ experience in applying statistics to agriculture, horticulture, biosecurity, entomology, ecology, plant breeding and finance. He has been an author of Genstat for 27 years developing the spreadsheet, statistical menus, client interface and graphics and has contributed over 80 Genstat procedures in areas of data manipulation, experimental design, probability distributions, model fitting, data mining and microarrays.