Data and Dressage

As someone who is pretty obsessed with data in my non-horse life, when data and horses come together, I just about die. When DRESSAGE and data come together, I just. I. I don't...

http://www.dressagedetective.com/

The coolest website ever. You can look up dressage judges and see what their average scoring was, the analyst also ranked every judge by generosity and popularity. Look up Axel Steiner and feel bad for anyone who has to ride in front of him (he's an excellent judge but look at the generosity ranking!).

The facebook page has a lot of cool stuff posted (from the website)- such as the most popular breeds at each level and finding out that Friesians score better than Swedish Warmbloods at First level.


This is so cool, and this is also time to talk about my horse related data science project.

Enjoy some data cartoons

I decided to do a side project because I enjoy being too busy to do anything I needed to bulk up my portfolio a bit. It had to be in data science, specifically using R to run my analyses. I couldn't come up with a project that sounded very fun and then I remembered...

Heidi and I were talking one day about school and I said that I loved statistics because I'm a nerd and she said she loved statistics too when she was in college! She described a project she did- taking information about the horses that she and her mom sold and creating a formula of sorts to decide how much the horses were worth, taking into account age, breed, level of training, show record, etc.

I was like oh so using a stepwise MRC to create a regression formula? YES SO COOL OMG.

After drooling over that project idea, I decided to replicate it, but using data from dreamhorse.com. Dreamhorse gives you age, breed, gender, height, temperament, and price, and the description usually states level of training, level of showing, and who the horse is suitable for. YES dreamhorse isn't actually an accurate depiction of what horses are selling for and instead I can only look at what horses are being priced at but I don't have access to that kind of data.

So I'm in data collection right now, I have 55 horses in a CSV file. I'm writing my R script as I go so I can quickly spit out descriptives and a few MRCs (I'm treating it like exploratory data analysis so I'm just fucking around a bit). My end goal is to determine what factors (as in age, breed, height, level of showing, etc) affect the horse's price. I can analyze the variance explained by each factor, giving me a strength of sorts for each factor.

Obviously we can speculate what the results may say- it is likely that the level of showing/training the horse has explains quite a bit of the variance in price with the higher level horses being more expensive. It's also likely that warmbloods might be more expensive than the other breeds.

He sure looks expensive

I can do really cool things too, like hold some factors constant. Meaning that I can do this: if we don't take breed into account (meaning, we statistically hold breed constant for all horses, changing the numbers to make it as if all horses were the same breed), how does that change what factors really matter in how people price horses? As in, if all horses were the same breed, would say level of showing matter MORE than if horses were different breeds?

The interesting thing about MRC is that you can also take into account overlap. For example, I want to see how much overlap there is between breed and temperament. I bet there's a relationship between the hotter breeds and a higher number under temperament. I can quantify that relationship.

Once as I have all my data (currently learning how to get python to scrape dreamhorse for me so data collection goes faster), I can start messing around even more. Right now is all about gathering as much data as I can.

Yep

Anyway, I think it's pretty cool and it has kept me relatively busy.

Don't judge me on my nerdiness, it has got to be contagious. My entire house (me, boyfriend, and two roommates, all of us in the same grad program) is running a statistics project using a game that we play regularly (Settlers of Catan, highly recommended even if you don't obsess over statistics while playing it). Also boyfriend and I have been talking briefly about doing some text analysis with twitter posts. I'm not the only one! I am, however, the only one sitting downstairs arguing with powershell while I enjoy some tea.


This afternoon I'm going to see if I can ride little TC again, and on Friday Ali wants me to ride him on a trail ride with her off the property! I'm excited, although it has been a while since I rode such a greenie out and about (although TC has had a lot of trail experience so it should be fine). Looking forward to it, although Rico is for sure my favorite trail horse.

The best!

Speaking of, I haven't heard anything from my mom about Rico's back feet, so I assume that he's doing fine! I guess I'll find out today since I will need to stop by the ranch to grab my saddle.