SOCR Analyses Expansion

SOCR Project - SOCR Analyses Expansion Project

Project Goal

To expand, and redesign as necessary, the current collection of SOCR Analyses Tools. This may involve implementing new SOCR analyses applets, redesigning the analyses as HTML5/JavaScript, refactoring and improving the organization of the SOCR Analyses package, extend the command-line SOCR analysis interface, etc.

Background

Explore and research the SOCR Analyses first. The Analyses package is one of the major SOCR packages and consists of a collection of applets that enables the statistical modeling, analysis and inference on user provided data. It currently has parametric and non-parametric, linear and non-linear, quantitative and qualitative, univariate and multivariate types of analyses.

Project specs

To expand the framework of SOCR analyses, first review the following resources:

SOCR SLR Randomization Test on Slope

This project aims to generate a new SOCR Simple Linear Regression (SLR) Analysis module (just replicate the SLR Analysis) that provides one additional functionality – permutation-based test on the slope (beta) coefficient.

This essentially will sample 1,000 to 10,000 times (user controlled) the data and estimate (call SLR) the beta-coefficients. Then compute the proportions of beta-coefficients that are (say) bigger than zero, and use their sampling distribution to infer the probability that the initially computed beta is statistically significantly non-trivial ($\beta \not= 0$).

Here are some examples, generated using R, that we are trying to implement as a new permutation-based SOCR SLR coefficient test.

Also see the SOCR Resampling and simulation webapp, and the corresponding Resampling project page, which does something similar – multiple simulations and inference based on permutations.

Adding a Hierarchical Clustering and Classification Analysis

Consider adding a new SOCR Hierarchical Data Clustering and Classification analysis applet. Below are some useful resources:

Adding the Skillings-Mack (SM) Analysis of Treatment Effects for Unbalanced Designs

The *Skillings-Mack* (SM) test-statistics provides a general method for comparing treatments for incomplete blocks when the observations may be randomly missing and the number of observations per cell are different. The SM) test-statistics generalized the Friedman test when there are missing data or there are different number of observations per cell. The resources below provide details and examples of how to compute the SM statistics and apply it to analyze real data.

Command-line SOCR analysis interface

To extend the command-line SOCR analysis interface, see the following resources:

References

The following references would be useful for this project: