There should be only one ingredient in a jar of honey, but sometimes, manufacturers like to get creative. Therefore, in 2018-2019, the Canadian Food Inspection Agency (CFIA) sampled 240 random jars of honey from across the country, and checked whether they exceed the expected threshold for non-honey sugar. The publicly-available results comprise dataset I will be using for this project, which includes information about the province in which the sample was taken, the country of origin for the product, the floral source of the honey, the amount of non-honey sugars, and whether or not the sample was judged as satisfactory, according to the CFIA (https://open.canada.ca/data/en/dataset/7ecfee4c-b6f5-45f8-9fd4-16976f55f8d9).
This will be an exploratory analysis with the goal of determining if changes in province of sampling, country of origin, or floral source increase the likelihood of an unsatisfactory rating. The ideal model will be a logistic regression with the rating as the outcome variable and province of sampling, country of origin, and floral source as predictor variables, but the specific structure will depend on what the data look like. A generalized linear regression model with the same effects structure could be run on the content of non-honey sugars (coded as C4 content in the dataset), however, the satisfactory/unsatisfactory outcome variable was chosen due to a lack of specialized knowledge about honey testing and how much C4 content is permitted in a sample before it is deemed unsatisfactory.
honey.rmd is where code was written, honey.md displays the same text, code, output, and images in a much more pleasant format. honey.pdf displays the same content as the website, but in PDF format. The dataset is available in honey.csv. This is a lightly edited version of a final assignment for LING 518 with Márton Sóskuthy at UBC in Fall 2021.