
Writing a data management plan: data transformations
There are a number of reasons you might need to transform your data during your research project or afterwards. Unlike the earlier discussion about migrating your file format, data transformation involves computing new values from old in the actual data content.
Some transformations are purely statistical, and are used to prepare the data to fit a model. An example was given on the previous page of one kind of transformation - logarithmic transformation to make a skewed distribution fit on a normal (bell) curve. Certain statistical techniques, such as linear regression for example, require data normalisation or some kind of transformation to begin analysis.
Of course there are implications for decisions about which form of the data to keep for the long term, which form to share with other researchers, and how to document all changes to the data. Other units in this module cover these topics.