We've been busy making lots of improvements to Analyse-it recently, so we thought we'd highlight the significant new and improved features.
If you have active maintenance, you can download and install the update now, see updating the software. If maintenance on your license has expired, you can renew it to get this update and forthcoming updates, see renew maintenance.
Fit Model now supports Probit regression. Probit regression is similar to logistic regression, as both use a link function to transform a linear model into a nonlinear relationship. A linear model uses the equation Y = α + β x, whereas both logit and probit equation use the form Y = f(α + β x). They only differ in the definition of the link function f(): the logit model uses the cumulative distribution function of the logistic distribution; the probit model uses the cumulative distribution function of the standard normal distribution. Both functions give a predicted probability, Y.
Health sciences, such as epidemiology, often use the logit model as the predictor coefficients are interpretable in terms of log odds-ratios. The probit model coefficients cannot be interpreted as easily but may produce a better fitting model in other scenarios. For example, in method validation, probit regression is used to model the hit rate of a molecular test. You can then use the model to establish a detection limit or determine diagnostic cut-off points from an underlying continuous response. For a guided example, see our blog post Calculating the detection limit for a SARS-CoV-2 RT-PCR test.
You can now apply a transformation to a variable during analysis. This feature is currently available on the Distribution and Fit Model (simple regressions) analysis, but it will be available in all analyses soon. To transform a variable, click the properties icon next to the variable selector drop-down, choose Transform, and then select the transformation function.
A lot of customers buy Analyse-it for ROC analysis, as it has always lead the way in diagnostic test analysis (https://pubmed.ncbi.nlm.nih.gov/12600955/). We recently extended the Binary (Sensitivity/Specificity) test to allow testing equivalence and non-inferiority hypotheses tests. And, now you can calculate predictive values for different population prevalences – ideal for modelling the behavior of a test in different scenarios.
Qualitative method comparison is now more prominent on the Method Comparison command menu, with clearer titles: Binary and Semi-Quantitative. We added Average agreement measures, which are useful when there is no reference/comparative method (for example when comparing laboratories or observers). There is also an excellent new plot for visualizing agreement between qualitative methods: the Bangdiwala agreement plot.
Finally, we released a new video tutorial for the CLSI EP15-A3 protocol.
We're keen to hear what other tutorials you would like to see for CLSI EP protocols. Contact us to let us know what you would like to see covered, and we'll put together more tutorials over the next few months.
For more information, see the online documentation: Estimating agreement between two binary or semi-quantitative methods Agreement plot Fitting a simple probit regression Comparing the accuracy of two binary diagnostic tests
Comments are now closed.
