Assignment 5 - Health Inspection Data

Publish to GitHub by Thursday June 2nd, 11:55pm. Hand in one printed copy for the whole group in class or office hours by June 3rd.

Your group assignment can be found here.

Start immediately and ask questions! Useful R functions: levels(), melt(), merge(), grep(), table()

Introduction

Health inspectors conduct routine surprise inspections at restaurants throughout the United States to ensure the safety of consumers. If minor health code violations are found, the restaurant is given a few days to correct them and a follow-up inspection is scheduled. If major violations are found, the restaurant may be temporarily or permanently closed.

Health inspections are typically managed by county or city governments. Public records are kept for every inspection. Most counties make these available for free online. Unfortunately, there's no official standard for how health inspection records should be organized, so it can be difficult to combine and analyze data from multiple counties.

Yelp developed the local inspector value-entry specification (LIVES) as a common standard for health inspection records. The LIVES format splits the records across three or more CSV files with specific columns.

In this assignment, your group will explore health inspection records for three counties in Northern California: Alameda, San Francisco, and Yolo. Most of the records have been converted to the LIVES format to aid analysis. See the link above for a full description of the format. The feed_info file has been omitted for all counties since it isn't particularly relevant.

Nonetheless, there are some differences in formatting between counties. You may have to do some light data cleaning to standardize the columns before analysis. Moreover, the city of Berkeley manages its health inspections separately from the rest of Alameda county and does not yet use the LIVES format. Part 1 will have you convert the Berkeley data to something similar to the LIVES format.

The goal of this assignment is to produce a high-quality report on a research question about the health inspection data. Don't let the word "research" intimidate you! All this means is that your group will choose a specific question to investigate in detail.

Part 1 - Cleaning The Berkeley Data

The file berkeley.csv contains the most recently published records for Berkeley. Each row corresponds to one inspection. Columns are used to indicate which violations were found.

1. The Berkeley data doesn't include a "score" column for the inspections. Compute a score for each inspection using the formula

   score = 100 - 4 * (# of minor violations) - 8 * (# of major violations)
   

   which is similar to the formula used in San Francisco.

2. Use the reshape2 package to reshape the Berkeley data so that each row corresponds to one violation. Name the new column "violation".

3. Commit your R script for cleaning the Berkeley data as berkeley.R. Save the cleaned data with saveRDS() and commit it as berkeley.rds.

Part 2 - Research Question

Come up with a research question about the health inspection data. Your question should be broad enough that you can come up with multiple strategies for investigation or follow up questions. Although there is no written work for this step, it is very important and you should spend some time together brainstorming and discussing. Consider:

• Why is the question interesting?

• Do you have the data necessary to answer the question in detail?

• How deep is the question? Are there multiple aspects to investigate?

• How can different aspects of the question be divided up among the group? Lay this out ahead of time.

Keep in mind that the data can be viewed from many different perspectives. For instance, your question does not necessarily need to concern health inspections and could simply use the data as a listing of restaurants. However, you must use data from all three counties (including Berkeley). Questions about a specific region or city can be expanded by making comparisons with other regions or cities.

It is okay and encouraged to supplement the health inspection data with other data sets. For example, you might want to use census data or the shapefiles that were provided with Assignment 4.

Feel free to consult Nick or the TAs if you're unsure whether your group's question is appropriate.

Part 3 - Analysis

Write a 4-8 page report (including graphics, but excluding code) with the following components:

• An introduction which includes a brief description of the data and explains your research question. Justify why your question is interesting and outline how you plan to investigate.

• An analysis which examines several different aspects of the question or examines the question from several different angles. Every group member must contribute to the analysis. Use a variety of graphics based on the data as evidence to support your claims. This should be the longest component of your report.

• A conclusion which synthesizes your analysis into a coherent answer to your research question. Explain the practical meaning and limitations of your answer. Also discuss how your answer could be expanded on if you had additional time and data.

Since the work will be divided among your group, it's a good idea to meet after assembling the report to proofread for problems with organization and flow. The report should not read like 3 separate reports. The report will be graded according to the class rubric, with consideration of the group evaluation forms.

Save your report, including graphics, as a PDF file called report.pdf. Save your R script for analyzing the data as analysis.R.

Part 4 - Group Evaluation

After completing part 1-3, individually and privately fill out the Assignment 5 group evaluation form. The form will be posted to Piazza on June 2nd and may be completed any time before June 6th. The form will ask you to summarize the contributions each group member made to the assignment and take at most 30 minutes to fill out.