My most common advice to students engaged in interpreting data for an assignment/project is to look for four things:
- Find links
- Observe trends
- What patterns emerge (eg clusters)
- Make predictions
This is basic advice, but it can be a useful starting point when you first open a data file and wonder what to do with it. Brent Dykes, writing in Forbes.com, tells us Why Companies Must Close The Data Literacy Divide. He offers loads of advice to improve data literacy for all. Included in this advice is some on Data Interpretation, and suggests making the following types of observations:
- Trends: What direction is a trended metric heading (up, down, flat)?
- Patterns: What repeatable patterns or cycles are you seeing in the data (e.g., seasonality)?
- Gaps: Are there any obvious gaps or omissions in the dataset?
- Clusters: Are some values bunched closely together in certain areas?
- Skewness: Are values noticeably concentrated or skewed more to one side than another?
- Outliers: Is there a data point that is detached or far removed from the rest of the data points?
- Focus: Has something in the chart or table been emphasized to draw attention to it? Is it obvious why part of the data was highlighted?
- Noise: Is there any extraneous data included that detracts from the main message of the chart?
- Logical: Does the data help to answer a specific business question? Does the data support a proposed conclusion or argument?
Source: Dykes (2017)
These are really simple and great suggestions that I will now add to my shorter list. Students will naturally be curious about any dataset that they use, and many won't need a list such as above to get going. Nevertheless, Dykes' list will make a great starting point and will form that basis of a good assignment or project. Data Analysts/Scientists need to have the skills and tools to enable them to make the above different types of observations almost immediately upon opening a data file. While the list is aimed specifically at charts, I feel that it can be applied to any type of data.