Practical Data Analysis: Theory-driven, Theory-directed

Peter Grzybek


Once ‘visualization’ and ‘interpretation’ are in the center of this year’s Digital Humanities Conference in Tartu, it should be clear, from the very beginning, that visualization cannot be but a particular step in the process of data interpretation, which cannot replace interpretation itself: either visualization comes up in an early stage of a scientific process, in terms of some kind of explorative approach, or it summarizingly illustrates the results achieved before. From this perspective, the whole matter boils down to the crucial question of data interpretation, a concept which has been a key concept in the humanities for more than a hundred years.

In this workshop, some concrete attempts will be made to illustrate what it would mean to arrive at scientific interpretations in terms of explanation of data from various fields of the humanities. As a matter of fact, the concept of explanation has changed over the decades, if not centuries, starting from causal to probabilistic concepts, where scientific explanation is held to be impossible without theories and laws: understanding theories as systems of laws, and laws as hypotheses related to other hypotheses from the field, and corroborated on relevant empirical data, no scientific explanation is possible without theories and laws.

Since the explanatory interpretation of data in this understanding thus is bound to the formulation of hypotheses, which can be submitted to (statistical) testing, the workshop will start on theory-based and theory-directed hypothesis formulation, followed by data acquisition, preparation and analysis, hypothesis testing, and finally interpretation.

By way of an illustration, the workshop will concentrate on two chosen examples: the structures of prose rhythm in texts of different languages, on the one hand, and shot length in movies, on the other. Tools for these analyses will be (made) available for participants, who are kindly asked to pre-register, along with their email addresses, and some additional information as to their personal scholarly profiles.