M.F.O. (Max Felix Orson) Welz

Due to modern technology and the internet, data collection is nowadays easier than ever. Consequently, datasets are growing ever larger, both in terms of sample size and dimension. On the other hand, large datasets are more likely contain potentially harmful outliers that are in turn difficult to detect given the dimension of the data. The issue of detecting outliers exacerbates when the data is noncontinuous, such as rating-scale data. However, statistical literature on outliers in rating-scale data is scarce.

In this PhD track, we will (i) develop methods for identifying outliers in (potentially large) rating-scale datasets; (ii) develop fast and efficient statistical estimation methods that are robust to outliers in high-dimensional data, in particular in rating-scale-data; (iii) derive mathematical guarantees such as convergence properties and asymptotic theories for the developed methods. The overall goal of this PhD track is to contribute to the scarce literature on robustness in noncontinuous data. The methods that we will develop are attractive for social scientists and researchers/practitioners that work with rating-scale data, as our proposed methods will make their analyses more reliable and less prone to the adverse effects of outliers.