Multi-group analyses

Multi-group analysis is a more advanced feature of NeuroMiner where classification is used to distinguish between subjects from multiple groups (e.g., schizophrenia, bipolar, and depression subjects). During data input, these groups are simply defined within the labels – e.g., ’SCZ’, ’BP’, and ’DEP’ identifiers. Once this is defined, then an option will appear in the parameter template called ”Multi-class settings”. Selecting this option and then selecting the option to ”Train multi-class predictor” to ’yes’ will reveal another option to ”3 : Specify multi-class decision mechanism [ ]”. This option determines how NeuroMiner deals with the multiple groups (e.g., how it optimises the predictions) and is critical for the interpretation of results. For all the methods included here, it is important to note that NeuroMiner works by conducting multiple binary analy ses between each group pair (see Aly, 2005 for an overview of the main types introduced below). Selecting this option will reveal the following menu:

1 | Simple Pairwise Decoding (Maximum-Wins method)

2 | Error-correcting output codes

3 | Hierarchical One-vs-One method

Simple pairwise decoding

This is the simplest method. Given that NeuroMiner has conducted different pair-wise classifications between all the group pairs, this option will categorize an individual by their maximum score across all the classifiers. For example, in a three group problem, if an individual’s score is highest for the schizophrenia group in the comparison between schizophre nia and bipolar (i.e., as opposed to the SCZ-BP, or SCZ-DEP, or BP-DEP) then they will be classified in the schizophrenia group.

Error-correcting output codes

Error-correcting output codes (ECOC) is a meta-method that is used to solve multi-group classification problems that ap plies a voting scheme (i.e., an ensemble method) to decide upon the correct class. It works by assigning a binary code to each group and then binary functions are learned one for each bit position in the binary code strings (see Dietterich et al., 1995 and Dietterich and Bakiri, 1995 and Kinderman et al., 2000 and for an overview see Aly, 2005). Optimal results are obtained by maximally separating the codes using a distance metric, such as the Hamming distance.

Hierarchical One-vs-One Method

This technique arranges the classes into a tree and then at each node of the tree a binary classifier makes the discrimination between the different child class clusters (see Aly, 2005 and Wang & Casasent, 2009).