Skip to main content icon/video/no-internet
icon backReturn to Entries

Data Models

Data models, or models of data, are processed versions of data, typically prepared in such a way as to make the data usable as evidence. The term first gained prominence in a 1962 paper by the philosopher of science Patrick Suppes, and after a period of relative neglect has now become an area of active research. This entry summarizes some of the key distinctions and debates, including those regarding the ontology of data, the difference between raw data and data models, the purpose(s) of data modeling, the variety of data processing methods used in the production of data models, and how data models should be evaluated.

All STEM research—indeed all empirical inquiry—is heavily reliant on data collection, processing, and interpretation; hence data models are of foundational importance. Conceptualizing the research process as involving data modeling helps to clarify the relationship between data and the world, data and theories, and the role of modeling in the scientific process.

Ontology of Data and Data Models

One of the most basic theoretical questions one can ask about data concerns what kinds of things data are. Some researchers place no restrictions on which elements of the world can count as data. On this view, laboratory mice could themselves count as data. Other researchers define data as records of a process of inquiry involving a physical interaction between the researcher, measuring instrument, and the world: It is only the results of observations or measurements performed on the mice colony that count as data. On this latter view, data necessarily involve a level of abstraction. Data involve replacing the thing in the world with a number, a photograph, or a recorded description. Which view one adopts arguably has epistemic implications such as whether data are unproblematically given or whether they are subject to the sort of questions raised by the philosophy of observation, philosophy of experiment, and philosophy of metrology (measurement).

What kind of thing is a data model? When Suppes first introduced the notion of a data model, he conceptualized them as Tarskian or set-theoretic models. Although Tarskian models are prominent in mathematics, the more commonly used notion of model in science today treats models as representations of some target system. Some have argued that this representational notion of a data model is important for recognizing that data are about the world, and hence for making sense of their ability to serve an evidential role.

Distinguishing Raw Data, Data Models, and Models

Data models are typically contrasted with raw data, but it is debatable whether and how such a distinction can be meaningfully drawn. Raw data are sometimes defined to be the immediate output of an observation, measuring process, or instrument. Often, however, such immediate outputs are not yet in a form that is usable by scientists. This can occur for a wide variety of reasons: First, the data-collection process might involve some obvious errors resulting in a certain portion of bad data that should be excluded from the data set. Second, there could be a source of noise in the data signal that needs to be filtered out or subtracted. Third, the output of the instrument might be of a quantity (measurand) that is closely related, but not identical, to the data quantity of interest, and hence must be converted before being useful. Fourth, data are often discrete samples of a continuous quantity and hence need to be smoothed or interpolated. Fifth, data often need to be organized or ordered before patterns can emerge. The outcome of these various processes of data wrangling—that is, of correcting, converting, smoothing, organizing, and so on—is what we call a data model.

...

  • Loading...
locked icon

Sign in to access this content

Get a 30 day FREE TRIAL

  • Watch videos from a variety of sources bringing classroom topics to life
  • Read modern, diverse business cases
  • Explore hundreds of books and reference titles

Read next

More like this

Sage Recommends

We found other relevant content for you on other Sage platforms.

close

Have you created a personal profile? Login or create a profile so that you can save clips, playlists and searches

close
close
close Signed In to profile You are signed in as:
close
Logged in Institution Institution