Introducting pvalue.io

This software is designed to perform statistical analyses using data collected on an Excel or Text file. Specialized in medical statistics, it is used to verify (or more correctly to reject) hypotheses using frequent statistical tests used in the health field.

The objective of this software is to be easy to use for all those who do not have enough statistical knowledge to perform their own analyses, and also has an educational objective. We therefore had to make compromises on the software’s functionalities and methodological aspects. Thus, we have chosen to set the alpha risk at 5%, to present the confidence intervals at 95%, and only bilateral superiority tests. Interactions are difficult to interpret and cannot be calculated.

Pvalue.io mainly uses 3 open source tools:

  • R, the software and reference programming language for statistical analysis
  • The Shiny package, an extension of the R software allowing to build graphical interfaces with client/server communication
  • The simplestats package, an extension of the R software allowing more appropriate operations to be carried out in biomedical research, guaranteeing the transparency of the analyses carried out by pvalue.io

Available analyses

But don’t worry, you don’t need to search for which test to use first, as the software will automatically select the right test.

Features not available

Measures that do not return a p-value are not available; these include:

  • Sensitivity, specificity, negative and positive predictive values (NPV, PPV), with or without a ROC curve
  • Concordance computation to obtain a Cohen’s kappa
  • Computation of the sample size

Caution

Be extremely vigilant: it is not possible to perform tests on repeated measures, or with matched samples. This is the case if:

  • The endpoint is computed from several measures collected from the same patient at different times (for example, blood pressure of included patients is measured; then antihypertensive therapy is administered for 1 month and blood pressure is measured again, then the difference is calculated)
  • The endpoint was measured before and after between 2 groups who had a different intervention (identical to the previous example, with a group of patients who received the treatment, and a group who received the placebo)
  • We set up a case-control study

The use of such statistical methods requires a statistical expertise to avoid any mistakes.