Saturday, November 10, 2018

Notre Dame University - "The Shirt"

Guest Post - by Michael Lieff (@virgonomic)


Every year for the last 15 years, my neighbour, who is a die hard Fighting Irish fan, has planned a driving trip to Notre Dame University near South Bend, Indiana. I attended for the first time in 2017 and again in 2018. After a travel day, the first stop on the campus tour is the bookstore. In the lobby, they have a table with one style of short- and long-sleeve t-shirts. In 2017 "the shirt" was navy and it didn't really grab me.


However, in 2018 the shirt was kelly green which drew me in, as green is my favourite colour. I read the price tag and learned that "the shirt" is a student initiative and the proceeds go back into student activities and assistance. At $18 USD it was a no-brainer.

Once I had my shirt, I visited the URL on the price tag. There is a link to a timeline that shows the shirt design from every year, and more importantly, the number of shirts sold, the team's record and the shirt manufacturer. Found data! Even more interesting is that there is no data for number sold for the years 1994-1996.

Analysis

The first question that came to my mind is: how many shirts did they sell from 1994-1996? Due to this gap, the dataset is a really nice example to explore interpolation and extrapolation. I figured the trend would be linear and the line of best fit would give a pretty logical prediction. Upon visualization, it definitely isn't cut-and-dried.

There are some interesting things going on here.The number of shirts sold dropped fairly significantly from 1993 to 1997. It also skyrocketed in 2002 and then plummeted in 2004. Possible reasons for this would make for an interesting discussion.

Drilling a bit deeper, the next question that came to mind is: Are more shirts sold in seasons where the team is winning?

It doesn't appear so, but I will let you 'do the math'.

Sample Questions

In terms of analysis, the following questions could be asked:
  • Is the trend linear or is a curve a better model?
  • Can you interpolate the number of shirts sold in 1994-1996 where there is missing data? Extrapolate the number sold in 2018 or beyond?
  • What are the mean, median and mode number sold?
  • Do the number of shirts sold correlate with the team’s wins that season?

Download the Data

 Let us know if you use this dataset or have any suggestions for things to do with it beyond this.

Monday, November 5, 2018

2018 NFL Salaries

We have a local NFL player that went to high school in one of the schools I support. Luke Willson was recently on the Seattle Seahawks and currently is on our local Detroit Lions. In conversation, a coworker wondered how much his salary was. The Internet provides. Not only his salary, but the salary of every one of the almost 1800 players (who knew there were so many?).

And when you have such a large data set, I think that you should analyze it. It's not a particularly deep topic. But it's a good data set to talk about mean, median, skewing and outliers. Not anything super interesting from a data perspective but the context may be interesting enough to capture the interest of some of your students to do basic single variable analysis. The data includes info about a player's name, salary, position, team, overall rank and I added the team rank. There are 32 teams and a bit over 50 players per team.

Analysis


Certainly some things you can do are to create some graphs. The first types that comes to mind is a dot plot, box plot and histogram. In this case the dot and box plot are provided by CODAP while the histogram comes from Google Sheets. You can see from the dot plot that the mean and median are quite separated (which we would expect from the skewing) and that there are a large number of outliers.

Since we were talking about Luke Willson, we could certainly ask how his salary compares to other NFL players (he's 455th) or other players on his team (he's 18th of 56) or even how he compares to other people the same position (21st of about 126 tight ends and is above the mean tight end salary)

Sample Questions

  • Determine the mean, median and standard deviation for the salaries attribute.
  • Which team has the highest mean salary? median salary?
  • Choose a player of your choice, how do they compare to the league, team and position?
  • Besides the way it looks, what confirms that this data is skewed to the right?
  • Which team has the highest number of outliers?

Download the Data

Let me know if you used this data set or if you have suggestions of what to do with it beyond this.

Friday, October 26, 2018

Walnut Crushing World Record

Check out this video (thanks to @ddmeyer for pointing this one out).

So the guy crushes walnuts with his head and what we get is a linear relationship. What I did was timed how long it took for each walnut to get crushed and collected in file (if you are interested, I slowed the video down by 505% then used an online timer to get the splits). So now you can do some analysis. It's not a particularly interesting data set but it might give a fun context to look at linear relationships.

Analysis

I guess the question that most comes to my mind (after "does he have a headache") is he crushing the walnuts at a constant rate. Careful observation might find a couple of spots where he hesitates a bit and you might want to discuss whether that shows up in the data. But is the data linear? Looking at the graph you can see that for the most part it is, but there is a slightly faster rate at the beginning and a slightly slower at the end but each section seems pretty linear.

Another thing you might want to discuss is whether it should be Time vs Walnuts or Walnuts vs Time. Since rates are usually per unit time then it probably makes sense to do Walnuts vs Time but you could argue that the total time depends on the number of walnuts or that the total number of walnuts you could crush depends on how much time you have. Note that the easiest way to swap the axes in a Google spreadsheet is by changing the position of the columns so to do that I just copied the Time column to both sides of the Walnuts column.

Sample Questions

Besides the above questions you could certainly ask:
  • What's the line of best fit?
  • What's the correlation?
  • How many walnuts do you think he could crush if it were two minutes? 10 minutes?
  • Is there a better fit than linear?
  • How many nuts would he have cracked if he kept at the same pace as the first 10 seconds?
  • If you only saw the first 5 seconds, what would be your prediction of the number crushed in 1 minute?
  • Can you tell, on the graph, when he hesitated?
  • What if he would have had the pace he finished with throughout the whole minute, how many nuts would he have cracked then? I think this was the previous record of 281

Download the Data

Let me know if you used this data set or if you have suggestions of what to do with it beyond this.

Sunday, August 26, 2018

Using the CODAP Online Statistics Software for Simple Analysis

So for years I have been a user of Fathom. Fathom is a dynamic statistical software package that has been available for teachers and students, free, here in Ontario. However, the software itself has not been updated over time and currently won't even run on a relatively recently purchased Mac. Not to fear, some of the creators of Fathom have come together to create the Common Online Data Analysis Platform (CODAP).

And because it was created by the people who gave us Fathom, it has a lot of similarities in style and function. It's not quite exactly the same but the biggest advantage is that it resides online so you can assign data for students to analyze and they can do so on any platform (probably not on a small screen phone very easily but still technically possible).
But for simple analysis, it does almost all the same things that Fathom did. Categorical and numerical analysis, mean & median, dot plots, scatter plots, linear regression, moveable lines, sum of squares, box plot, outliers and more. Some things it doesn't do (yet) are make bar graphs (though it makes the equivalent with dot plots) and histograms (though this may become an added feature). You can watch how easy it is to do some of those things dealing with simple analysis on the video seen below. If you want to play along with the video, here is the file that I used.

Once you know how to use the app, getting the data to your students is the next step. My preference is to have a pre-made CODAP file available for upload to CODAP. You can upload a file directly from any computer or conversely from a Google Drive. My preference is to do so from a Google drive. I have taken the liberty of converting many of the data sets on this blog to CODAP files. I have tagged all of them with the CODAP label here (also seen on the right side of the blog) or I have collected all the CODAP files in this folder. Conversely you can upload your own data in a .csv file. Though it does not seem like you can do this directly from a Google Drive. So I would stick to creating the CODAP files and sharing that with your students (either on Google drive or a local network drive). Either way, if you use any of these files, I would download them from this blog and then upload them to your preferred place.

And being redundant, here is a list of the past posts that I have done the conversion for and future posts will also have CODAP versions included.
Anscombe's Quartet
Smoking and Cancer
Movie Data
How Much Would you Pay for a $50 Gift Card?
Earthquake Data
Trending Data
Magazines
Speed Data
Electric Car Rebates
Is Levelling Up in Pokemon Go Exponential
Collecting Data from Pokemon Go

Don't forget to look at the CODAP site for lots of great resources. From more data sets, tutorials, FAQs and even though we haven't talked about them here, simulations. Or just look at the Educator Resources page.

Download the Data

All the Posts
Folder of CODAP files