ETC5521 Worksheet 1

Introduction to exploratory data analysis

Author

Prof. Di Cook

🎯 Objectives

The goal of this worksheet is to tackle a data analysis together, by

mapping out an analysis, with class input and help of AI
work on cleaning a data set
making some plots
discussing what is surprising, and what is expected

📋 About the data

The data to use is available from Tidy Tuesday 28 May 2024 page. Download the data from here, ideally using the tidytuesdayR package. You should only download the data from the Tidy Tuesday once, and save a copy locally on your computer.

In addition the gardenR package, available from remotes::install_github("llendway/gardenR") has extra details about the garden.

Solution

This is the code to download the data, and load relevant libraries.

Code

# remotes::install_github("llendway/gardenR")
# install.packages("tidytuesdayR")
library(tidytuesdayR)
tuesdata <- tidytuesdayR::tt_load('2024-05-28')
save(tuesdata, file="tuesdata.rda")

Code

library(gardenR)
library(tidyverse)
library(ggbeeswarm)

load("tuesdata.rda")
harvest_2020 <- tuesdata$harvest_2020
harvest_2021 <- tuesdata$harvest_2021
planting_2020 <- tuesdata$planting_2020
planting_2021 <- tuesdata$planting_2021
spending_2020 <- tuesdata$spending_2020
spending_2021 <- tuesdata$spending_2021

data("garden_coords")

🧩 Tasks

1. What are the variable types, how and when was the data collected?

Solution

This is collected in 2020 and 2021, and has a variety of numeric and categorical and time variables.

2. What would some possible questions be to ask about this data?

Solution

The link to the ChatGPT conversation is here. It was prompted by

Do you know the gardenR package by Lisa Lendway?
What might be some questions that we could answer with this data?

We chose to tackle one of the Economics questions: “How much produce (by weight or value) was harvested per dollar spent?” But realised that it was not possible answer this particular question with this data. It was refined to be:

Compare the ROI for varieties of beans.

The next step was to filter the data to focus on one vegetable, beans, and one year to get started.

Code

beans_planting_2020 <- planting_2020 |>
  filter(vegetable == "beans")
beans_harvest_2020 <- harvest_2020 |>
  filter(vegetable == "beans")
beans_spending_2020 <- spending_2020 |>
  filter(vegetable == "beans")

3. What do you expect to find?

Solution

We might expect some difference in ROI between varieties.

4. Pick one of the questions, and let’s try to answer it.

What summaries should we make?
Are we likely going to need to pre-process the data in any way?

Solution

We need to decide on a common scale. Steps are:

examine the price for each variety
summarise the planting data by variety and join to spending data.
summarise the harvest data by variety and join to spending.

Code

beans_spending_2020 |>
  select(variety, brand, price_with_tax)

# A tibble: 3 × 3
  variety             brand          price_with_tax
  <chr>               <chr>                   <dbl>
1 Bush Bush Slender   Renee's Garden           3.01
2 Chinese Red Noodle  Baker Creek              3.24
3 Classic Slenderette Renee's Garden           3.23

Code

beans_planting_2020_smry <- beans_planting_2020 |>
  group_by(variety) |>
  summarise(number_seeds_planted = sum(number_seeds_planted))

beans_2020_smry <- beans_spending_2020 |>
  select(variety, brand, price_with_tax) |>
  left_join(beans_planting_2020_smry)

ggplot(beans_2020_smry, 
       aes(price_with_tax, 
           number_seeds_planted)) + geom_point()

Code

beans_2020_smry |> 
  mutate(pr_per_seed = price_with_tax/number_seeds_planted) |>
  select(variety, pr_per_seed)

# A tibble: 3 × 2
  variety             pr_per_seed
  <chr>                     <dbl>
1 Bush Bush Slender        0.0752
2 Chinese Red Noodle       0.324 
3 Classic Slenderette      0.111

Code

# Do all the seed packs have the same number of seeds, 
# or did Lisa plant every seed in every pack?

# Harvest
beans_harvest_2020_smry <- beans_harvest_2020 |>
  group_by(variety) |>
  summarise(weight = sum(weight))
beans_harvest_2020_smry

# A tibble: 3 × 2
  variety             weight
  <chr>                <dbl>
1 Bush Bush Slender    10038
2 Chinese Red Noodle     356
3 Classic Slenderette   1635

Code

beans_2020_smry <- beans_2020_smry |>
  left_join(beans_harvest_2020_smry)

beans_2020_smry <- beans_2020_smry |>
  mutate(psy = weight/number_seeds_planted)
beans_2020_smry

# A tibble: 3 × 6
  variety             brand     price_with_tax number_seeds_planted weight   psy
  <chr>               <chr>              <dbl>                <dbl>  <dbl> <dbl>
1 Bush Bush Slender   Renee's …           3.01                   40  10038 251. 
2 Chinese Red Noodle  Baker Cr…           3.24                   10    356  35.6
3 Classic Slenderette Renee's …           3.23                   29   1635  56.4

Problems encountered.

Spending didn’t specify what the cost involved, whether it was one packet of seeds,
Not clear whether all seeds in each packet were planted. Numbers suggest, no, because some varieties had many sends and others very few.

Adjustments:

weight was adjusted by number of seeds planted

Conclusion:

Bush Bush Slender outperforms the other two, by a lot! This should also indicate that this variety is a better ROI.

Caveats: Need to check that

the plots where each was planted was a reasonable chance that they had same access to sunlight and nutrients.
Assuming that seeds were planted at same distance from each other.

Next steps

Examine the 2021 data. If same varieties grown do the same results happen.

Problems discovered in doing this: Name of variety in 2021 might have changed to be just “Bush”, and the other two were not used. Could compare against the one new variety.