Failed with gusto

Failing with gusto

This week I decided to try and participate in hacking open data in Indianapolis, way before I realized how tired and exhausted the Interventional Radiology week would be. I have been working on the Data Science course from Coursera with the plan that I should become a proper data scientist. I had a biostatistics class in the Informatics master’s program but the professor felt the best way to learn was to calculate all the problems only on a calculator. While that is great for PHD students in biostatistics, this left me unable to analyze big datasets on my own using statistical packages, and in turn meant I could not play in the big data game. I have R skills that I have picked up with my projects, but I thought proper training would improve my analytic skills and make me critical to data centered operations. Moreover, I have been working on registry systems to collect /aggregate data and analyze it, hence superior data skills are no longer a luxury but mission critical.

By yesterday evening I had moved on from the idea of participating, having left the hospital at 8 am and just exhausted. I slept in, but around 11.30 AM checked the site and with Nicole’s encouragement, I decided to try and hack on the Indiana State Auditor data. This meant i had less than 6 hours to develop an application. See competition details here . The competition rules are very unfriendly, requiring giving up copyright for the applications developed, and a commitment to work on the application with no form of support for the next 9 days. Nonetheless I started to look at the available data set from the auditor office that provided all credit card transactions for the last year(2015-2016).

I worked on an application based on R studio, R and shiny to develop an interactive dashboard that allows you to drill into the data and search by merchant, employee, cost paid, postal code and transaction IDs. See the hosted application here: https://gichoya.shinyapps.io/HackApp/

Data dashboard

Image showing the home page of the data dashboard

Despite the effort, my application crashed during demonstration time, and couldn’t show it off. Turns out that R did not like my use of the keyword ‘data’ and hence refused to load the dashboard after pushing code to github. My partner in crime aka Nicole did mention that my lack of emphasis on the low value of graphs to represent data while presenting to the judges was shooting myself in the foot. In my defense , I have seen multiple open data projects and visualizations are great, but in truth are hardly understood by most people. There is more value in providing resources to answer a specific question, which is what I tried to solve by allowing users to drill in and filter data that answers a specific question.

Ladies and gentlemen, that’s how to fail a Hackathon or better yet ‘Fail with gusto’

Code repo:

https://github.com/judywawira/IndyHack