2015 Data Science Salary Survey

Tools and Salary: A More Complete Model

WE ARE NOW READY to incorporate tools into a third salary model. We keep the same pool of features available as in the second model, plus one feature for each tool, and also keep the same subsample (no professors, students, or management). The larger clusters in the 2014 report were more conducive to being converted into features (as the number of tools in a given cluster that someone uses), but here it makes more sense to keep the tool-features as binary variables representing the usage/non-usage of one tool.

In addition to tools, we also add two features for cloud computing: one for the amount of cloud computing, the other for the type of cloud computing (public or private; this feature turns out to be insignificant in the model).

Most of the features kept in the previous model remain, and eleven tools are now included. The R² has only modestly increased, to 0.427.

26393 intercept +1505 age (per year of age above 18) +6106 bargaining skills (times 1 for "poor" skillsto 5 for "excellent" skills)  +420 work_week (times # hours in week) -2785 gender=Female +3012 industry=Software (incl. security, cloud services) -6412 industry=Education +1412 company size: 2500+ +9274 PhD  +919 master's degree (but no PhD)  +101 academic specialty in computer science+14667 California+10693 Northeast US  +231 Southern US  -451 Canada -1486 UK/Ireland-17084 Europe (except UK/I)-21077 Latin America-26146 Asia+8489 Meetings: 1 - 3 hours / day+9461 Meetings: 4+ ...