Multivariate Analysis
Exploration of complex associations between multiple variables using linear regression, principal component analysis, or simple models that allow controlling for confounding factors.
What it is
Multivariate analysis is the most advanced stage of statistical processing. It studies phenomena where several variables interact at the same time.
We use multiple regression (linear, logistic or Cox) to adjust results for possible biases, principal component analysis to simplify complex data and cluster analysis to find groups, and we explain the results in the context of your research.
What's included
- Multiple regressionLinear, logistic or Cox, adjusting for confounders.
- Principal component analysisSimplifies datasets with many variables.
- Cluster analysisReveals hidden groups in your sample.
- Choice of techniqueWe assess your sample and your statistical power to choose the model that fits.
- InterpretationClear and technical, in the context of your research.
What you receive
- Modeling reportPDFThe reasons for the chosen models and a detailed interpretation.
- Coefficient and fit tablesExcel · CSVCoefficients, explained variance, p-values and goodness-of-fit statistics.
- VisualizationsJPG · TIFGraphics with legends and captions, for presentations or publications.
- Materials and Methods sectionThe procedures, the software and the criteria used to choose the models.
Frequently asked questions
When do I need a multivariate analysis?
When you want to know how several factors act together on an outcome, or when you need to adjust a comparison for confounders. Comparing one variable at a time cannot do that.
Which regression do I need?
It depends on the outcome you measure: linear regression for a continuous outcome (such as blood pressure), logistic regression for a yes-or-no outcome (such as disease present or absent) and Cox models for time-to-event outcomes (survival). All of them can be adjusted for confounders.
Do I need a very large sample?
It depends on the number of variables. We assess the statistical power and recommend the most suitable technique.
Tell us about your study
We reply with questions and a quote.