Business data induction 2023–02–20
What is panel data and how do we use it?
First ask what do we mean by panel data? Data in a panel, look up a definition of panel.
What kind of data do we consider in our team?
- Change in interest rates, values, revenue, number of staff, expenses, profit.
- Data range: year period, 5 year, calendar, monthly… regular
- Which company, industry average, market/type.
Typically we have multiple of these three kinds of things.
We can show two dimensions easily in a table, one on each axis (rows and columns). So that might be revenue against year for one company. If you have multiple companies, you will have to put that data directly below or to the right, with a row/column to identify the company. These blocks for each company are called panels.
Students may approach us with a handful of companies, months, items. Or perhaps 1,000 companies with 30 items over 30 years. You can apply different techniques at different scales, you need to automate at larger scale (typically PhD and above).
We looked at an example in our own area, enquiries statistics for the library. Here, we could could put month on each row, and service area in each column, with the cells showing number of enquiries per service area per month.
If you want to study month by month, or team by team, you can do that easily in the table and make an easy chart. However, if you want to study year by year, you would have to count 12 months at a time, and each of these 12-month groups would be a panel.
How do I work with panel data?
Different databases provide different options with panel structure. Sometimes you have a lot of control, sometimes you don’t. The term to change the panel structure is ‘reshape’.
The library (Data Specialist) offers limited one-to-one help with reshaping data:
- We point to one FAQ on reshaping in Stata.
- With Datastream, we can advise people on how to build a Request Table to produce a custom panel structure.
- Sometimes we can advise on building custom formula structures in Bloomberg, Eikon and Capital IQ Excel add-in enquiries.