2015 Data Science Salary Survey

Integrating Job Titles into Our Final Model

THE OMISSION OF JOB TITLES as features in the models we’ve so far presented is deliberate: we want to see how much can be predicated only from demographics and information about what someone does, not what they are called. This also allows us to compare the model without titles to a fourth and final model with titles, to see if job titles give us information not extractible from the other data we have about each individual. Before we show this model, it is worth describing the job title categories we are using in the context of the other variables we have been working with: demographics, tasks, and tools. As with the second and third models, we will restrict this section to the non-managerial and non-academic groups.

Classifying Job Titles

Respondents entered their job titles into a text field (as opposed to picking a choice from a drop-down menu), and we have classified the entries using a few simple rules to remove the overlapping respondents who would otherwise qualify for more than one group.^[10]

Little can be said with any certainty about some of the smaller groups such as DBA and statistician: the former tends to use Perl and work for older companies, the latter tends to use R and not use any cloud computing, but none of these observations are backed up by much statistical significance. However, these titles do not appear to be very common in the space, and we would expect that many who could call themselves statisticians could ...