Localising the auditory N1m with event-related beamformers:localisation accuracy following bilateral and unilateral stimulation

The auditory evoked N1m-P2m response complex presents a challenging case for MEG source-modelling, because symmetrical, phase-locked activity occurs in the hemispheres both contralateral and ipsilateral to stimulation. Beamformer methods, in particular, can be susceptible to localisation bias and spurious sources under these conditions. This study explored the accuracy and efficiency of event-related beamformer source models for auditory MEG data under typical experimental conditions: monaural and diotic stimulation; and whole-head beamformer analysis compared to a half-head analysis using only sensors from the hemisphere contralateral to stimulation. Event-related beamformer localisations were also compared with more traditional single-dipole models. At the group level, the event-related beamformer performed equally well as the single-dipole models in terms of accuracy for both the N1m and the P2m, and in terms of efficiency (number of successful source models) for the N1m. The results yielded by the half-head analysis did not differ significantly from those produced by the traditional whole-head analysis. Any localisation bias caused by the presence of correlated sources is minimal in the context of the inter-individual variability in source localisations. In conclusion, event-related beamformers provide a useful alternative to equivalent-current dipole models in localisation of auditory evoked responses.

The sensitivity of mapping methods to reference data quality:training supervised image classifications with imperfect reference data

The accuracy of a map is dependent on the reference dataset used in its construction. Classification analyses used in thematic mapping can, for example, be sensitive to a range of sampling and data quality concerns. With particular focus on the latter, the effects of reference data quality on land cover classifications from airborne thematic mapper data are explored. Variations in sampling intensity and effort are highlighted in a dataset that is widely used in mapping and modelling studies; these may need accounting for in analyses. The quality of the labelling in the reference dataset was also a key variable influencing mapping accuracy. Accuracy varied with the amount and nature of mislabelled training cases with the nature of the effects varying between classifiers. The largest impacts on accuracy occurred when mislabelling involved confusion between similar classes. Accuracy was also typically negatively related to the magnitude of mislabelled cases and the support vector machine (SVM), which has been claimed to be relatively insensitive to training data error, was the most sensitive of the set of classifiers investigated, with overall classification accuracy declining by 8% (significant at 95% level of confidence) with the use of a training set containing 20% mislabelled cases.

A novel ranking procedure for forecasting approaches using Data Envelopment Analysis

To compare the accuracy of different forecasting approaches an error measure is required. Many error measures have been proposed in the literature, however in practice there are some situations where different measures yield different decisions on forecasting approach selection and there is no agreement on which approach should be used. Generally forecasting measures represent ratios or percentages providing an overall image of how well fitted the forecasting technique is to the observations. This paper proposes a multiplicative Data Envelopment Analysis (DEA) model in order to rank several forecasting techniques. We demonstrate the proposed model by applying it to the set of yearly time series of the M3 competition. The usefulness of the proposed approach has been tested using the M3-competition where five error measures have been applied in and aggregated to a single DEA score.