Embedded Files
The first step, once data is available, is to conduct Exploratory Data Analysis or EDA. There are multiple reasons to carry out an EDA:
to check the data file contains the data you expect;
to explore some basic summaries and plots of the data;
to look for outliers or unexpected behaviour;
to help guide further analysis and interpret future results.
An EDA in its most basic sense is about visualising the data, and many of the business intelligence systems simply provide dashboards of EDA results. EDA can be revealing enough that further analysis is unnecessary but even when it is known that further analysis is to be carried out the EDA should still be carried out and can be a valuable complement to further analysis.
At the AMRC we have developed some guidance to carrying out an EDA, from what you are looking for to how to achieve it, and finally how to present it.
To the right, is a summary of the types of questions that are aimed to be answered through the EDA. Below are links to further guidance in producing and interpreting the results of the EDA.
Finally, for AMRC staff, we have produced a template report that can be used to record the results of the EDA.
For further support with conducting or presenting an EDA...
For further support with conducting or presenting an EDA...
An exploratory data analysis usually takes the form of summary statistics or plots that can explain basic features of the data, such as its shape, its centre and its spread and any simple relationships that exist. The results of the EDA should be shared amongst stakeholders for initial interpretation and to ensure that the data is suitable to meet the requirements of the business case and thus the project is still viable (or even necessary).
A Workflow for Exploratory Data Analysis
A Workflow for Exploratory Data Analysis
Using the workflow - reporting on EDA
Using the workflow - reporting on EDA
Data Reports
Data Reports
Tell the story of your datasets so they can be thoroughly and easily understood by customers and colleagues. This reporting template distills the process down so you can record the important points during the analytical process.
What do the summary statistics mean?
What do the summary statistics mean?
Summary statistics can reveal interesting features in the data and suggest what further analysis is appropriate.
What plots?
What plots?
Different plots can be used for different insights and be useful for interpretation of further analysis.
How to produce good plots?
How to produce good plots?
There are some simple rules to follow best practice when plotting the results of EDA.
Page updated
Report abuse