Know your data

Overview

Teaching: 5 min
Exercises: 5 min
Questions
  • What information is in our data?

  • What kinds of questions can our data answer?

  • What kinds of questions can not be answered by our data?

Objectives
  • Think about the data and it’s potential uses.

What the data says and what it doesn’t say

The data

Each row includes the following information:

  1. Branch ID
  2. Branch Name
  3. Number of Holds
  4. Title
  5. Author
  6. As of Date
  7. Web URL

Given a row like this one:

Branch ID Branch Name Number of Holds Title Author As of Date Web URL
EPLLON Londonderry Branch 36 The girl on the train / Paula Hawkins Hawkins Paula 03/16/2015 12:00:00 AM http://epl.bibliocommons.com/search?t=smart&q=the%20girl%20on

We can interpret this in words as “On the week preceding March 16, 2015, The girl on the train by Paula Hawkins was one of the top-ten most requested items at the Londonderry Branch. There were 36 holds for this item”.

Note that for this week and for this branch, there will be nine other rows containing the other top-ten titles.

Questions about the data

Think about the data as a collection, and contemplate the following questions:

Question

Notice that the above row does not give us enough information to tell us what position in the top-ten The girl on the train was at Londonderry Branch on March 16, 2015.

But can we figure this out from this data set?

Answer

Yes. In general to answer questions like this, we need to find all of the items in the top ten for that week at the Londonderry Branch (there should be 10 of them). We can then order by the number of holds to find out where The girl on the train sat that week. In our case, our data already has some organization to it, and we can answer this question by looking at the first ten rows of our data file.

Question

Is there any information in our data set that is either redundant or highly correlated?

Answer

Ideally, the Branch ID and the Branch Name fields should be pretty closely correlated.

The Title data also includes author information.

Probably the Web URL will hold similar information as Title.

Question

If I have a date and a title, can I add up the holds in the rows in the data that match to get an indication of the total number of holds for that item at that date for all of EPL?

Answer

Not really. This data is for those branches for which the title was in the top 10 holds. This will not include data for branches where the title was not in the top 10 titles, so we will not get the total number of holds.

Question

If I have a branch name and a title, can I add up the number of holds over the weeks to get the total number of people requesting that item at that branch?

Answer

No. Suppose you wait longer than a week for your hold to come, then you will be included in at least two different weeks worth of data, and you will be counted twice

Question

Can I use the data to determine how many weeks a title was in the top 10 at a particular branch?

Answer

Yes. You would look for the rows where the title and the branch match and count the number of rows.

Key Points

  • It’s important to think about what questions your data can answer.