• Check the assessment and project deadline map.
• Work from a topic to an available dataset and a possible research question.
• Discuss whether the data are suitable, using a few simple R checks.

Thinking about the last fortnight:

• How are you finding the balance between taught sessions and independent study?
• What has surprised you about the course so far?
• How are you keeping in touch with others on the course to share ideas, stay on track, or check anything you might have missed?

Take a moment to think, then share what you are comfortable sharing.

Think about a dataset search you have tried:

• Where did you look, and what search terms did you use?
• What helped you find a promising dataset?
• What was difficult, or made you reject a result?
• What would you recommend someone else try?

Let’s collect useful sources and search tips as a group. An unsuccessful search is useful to discuss too.

The reference slides include:

• Places to search: OSF, Zenodo, Kaggle, UK Data Service, and Rdatasets, with links and search tips.
• Your dataset checklist: assess your own candidate before committing.
• Keeping the scope manageable: possible simplifications and tidyverse functions.
• Kaggle checks: tracing sources, documentation, and reuse terms.
• Importing and inspecting data: file formats, R functions, paths, and simple checks.

Let’s explore Rdatasets together in the browser.

R demo: Explore the data and run a few simple checks.

library(tidyverse)
dat <- ggplot2::msleep
?msleep
glimpse(dat)
View(dat)

?msleep opens the dataset documentation; View(dat) opens the data viewer.

The documentation describes 83 species, including:

• sleep_total: total sleep in hours;
• vore: diet group;
• bodywt: body weight in kilograms.

The data include sleep duration and diet groups, so one possible question is:

• Outcome: sleep_total; grouping variable: vore.
• Explain why the comparison is interesting, using relevant background.
• Keep the claim within the design: this is a comparison, not evidence that diet causes sleep differences.

We developed the question from the available measurements. If patterns in the results later suggest a new question, describe that exploration honestly.

Start a new conversation without your fav AI and use this prompt. Enable web search if available.

Everyone submits the same prompt independently, then compares with a partner.

(Check the questions on the next slide!)

Open the suggested dataset and documentation in your browser.

• Did you get the same dataset? What differed in the answers?
• Can you find the actual data file and confirm the measurements and unit of observation?
• Do the source pages support the claims about access and reuse?

Share one verified detail and one claim that was wrong, unclear, or unsupported. A missing or broken link is a finding too.

Would you use this dataset, and what would you still need to check?

Choose a topic.

1. Find one candidate and open its documentation.
2. Identify what was measured and what one row represents.
3. Suggest one meaningful question those measurements could answer.
4. Use the checklist (and simple R checks) to assess the choice.
5. Share your decision and one unresolved issue.

What do you find difficult about finding datasets? What helped?

Share:

• a dataset or source you could use;
• a question grounded in its measurements;
• what you need to check or do next.

Raise any unresolved assessment or deadline questions.

If the question becomes a comparison of herbivores and carnivores:

# Choose columns
sleep_small <- select(dat, name, vore, sleep_total)
# Choose rows
sleep_small <- filter(sleep_small, vore == "herbi" | vore == "carni")

This leaves 51 species and excludes other/unknown diets. Explain the narrower scope; related species can still share evolutionary history.

You may not need to change the data. Consider a simpler source if preparation becomes too demanding.

Kaggle can be useful, especially for CSV files with previews.

Treat it as a starting point, not automatic approval.

A Kaggle dataset is only suitable if you can:

• identify the original source or documentation;
• understand the column names and values;
• find a usable outcome and predictors;
• explain a behavioural or psychological question;
• check that reuse is allowed.

Use the right function to import your dataset.

library(tidyverse)
library(readxl)
library(haven)

ldt_csv <- read_csv("data/chinese_ldt.csv")
ldt_excel <- read_excel("data/chinese_ldt.xlsx")
ldt_spss <- read_sav("data/chinese_ldt.sav")

Check that you can import and inspect the data.

After importing, do not start modelling immediately.

glimpse(ldt_csv)
names(ldt_csv)
count(ldt_csv, lexicality)
count(ldt_csv, script, lexicality)

• Are the variables named clearly?
• Do categories mean what you think they mean?
• Are numeric variables really numeric?
• Are missing values coded sensibly?
• Does the imported dataset match the documentation?