RELATIONAL STRATEGIES IN STATISTICAL DATA ANALYSIS: MAPPING, INFERENCES, CLUSTERING, TOPOLOGIES, MATCHING

Modelling reality is representing phenomena: descriptions of events are not always reducible to rigorous mathematical functions, requesting therefore statistical methods to reproduce their effects. Throughout both classical and innovative procedures of statistical analysis, the implementation of software can classify data, improving results with every step of sequence (mapping, inferences, clustering, topologies, matching). Assuming a critic opinion can only descend by a probabilistic choose, multiple tests formulize hypothesis on nature of data: this method of statistical inference is implemented in post_analysis.exe to map a set of raw spot data in 3D, to find stationary points and to identify possible outliers applying a multivariate analysis on results. Then, cluster_analysis.exe can group data using two iterative cycles which control clusters with parameters fixed by users. When geometry is unavailable in data, some spatial references are often provided by relationships of presence, association and proximity between elements of the set: therefore, graphs ' theory helps in organizing and managing this information. From data grouped in 3D, texture_analysis.exe builds up symbolic matrices and creates a structure for every cluster provided; the aim of optimize memory occupation is then fulfilled by new_sorting.exe, which rearrange and dissect graphs. The problem of matching searches for corresponding characteristics between different descriptions: with a digression in the Psychology's sphere, use of perceptual organization leads the implementation of software. To recognize similarity in comparable graphs, exact_matching.exe and unexact_matching.exe generate an identification code for every point, classifying nodes to associat