Consider who might create the data you are looking for. Could it be produced by the government? An advocacy organization? Perhaps a business or an academic researcher?
Remember the differences between quantitative and qualitative data. Quantitative data (i.e., entries are assigned numerical values) can answer questions about how often or how much a phenomenon occurs; this data can also be readily evaluated with statistics or used to create a mathematical model. Qualitative data (i.e., non-numerical and typically collected through observation or interviews) can be used to describe and characterize a phenomenon.
Be conscious of how data is collected. Cross-sectional data examines a subject at a particular moment in time, while longitudinal data monitors change over time. Knowing how data was collected can tip you off to potential biases in the set.
All 15 executive departments of the U.S. federal government collect and publish data. Those 15 departments are:
Agriculture | Commerce | Defense |
Education | Energy | Health & Human Services |
Homeland Security | Housing & Human Development | Interior |
Justice | Labor | State |
Transportation | Treasury | Veteran Affairs |
Other federal agencies and commission may publish data of their own as well.
Various nonprofit organization produce and publish data, typically to inform viewers of the subject of their operations. What nonprofits will have useful data will depend on your field of study. For instance, a medical student would want to search for datasets published by the American Red Cross, the Kaiser Family Foundation, or the American Heart Association.
Corporations and business analysts will produce their own datasets to evaluate a company's performance or analyze trends within an industry. Some of these reports can be found through Business Source Ultimate (see below), but in general they are quite expensive and it is difficult for the library to acquire them, even through interlibrary loan.
Scholars in many fields produce data during their research. Some authors of scholarly articles will publish these datasets alongside the article, and others are willing to share the datasets if contacted by email. Some individuals will also upload datasets to repositories such as Figshare and DataONE (see below).