Bio302 - Biostatistikk – Data lifecycle

Why do biologists need statistics anyway?

Why do YOU need

data handling skills
data visualisation skills
statistics skills

Biology is applied data science!

Nearly all papers include data visualisation and analysis
Need to be able to understand the methods used to evaluate paper
Need to be able to make visualisations and analyses for papers, reports, theses

Changing expectations

FAIR and open science
Reproducible research

Most scientists ‘can’t replicate studies by their peers’

Data life-cycle

Experimental Design

Lecture 5

Observational studies vs experiments
Randomisation
How many replicates do you need?
Statistical traps

Data entry

Make your data easy to import.
Proof read & data validation
Lock data file. No more edits.

Spreadsheets

Practical recommendations for organizing spreadsheet data to reduce errors and ease later analyses.

be consistent
write dates as YYYY-MM-DD
do not leave any cells empty
put just one thing in a cell
organize the data as a single rectangle (rows = subjects; columns = variables; single header row)
create a data dictionary
do not include calculations in the raw data files
do not use font, colour or highlighting as data
choose good names for things
use data validation to avoid data entry errors
save the data in plain text files.

Examples

When is 1-9-2021?

Sex coded as 0 & 1

Data importing

R can import almost any data type.

CSV files with readr
Excel files with readxl
shapefiles with sf
NetCDF with ncdf4

Data cleaning

Process your data with code
Reproducible analyses
tidyverse (Lecture 3)

library(tidyverse)
data(penguins, package = "palmerpenguins")
penguins |>
  group_by(species) |>
  summarise(
    n = n(),
    mean_mass = mean(body_mass_g, na.rm = TRUE))

# A tibble: 3 × 3
  species       n mean_mass
  <fct>     <int>     <dbl>
1 Adelie      152     3701.
2 Chinstrap    68     3733.
3 Gentoo      124     5076.

Visualise with ggplot2

library(ggplot2)
library(ggbeeswarm)
ggplot(penguins, aes(x = species,
                     y = bill_length_mm,
                     colour = species)) +
  geom_beeswarm() +
  scale_colour_brewer(palette = "Set1") +
  labs(x = "Species", y = "Bill length mm", colour = "Species")

Data analysis

Huge number of statistical methods

This course will focus on

Exploratory data analysis (6)
Descriptive statistics (6)
Linear models (7--8)
Generalised linear models (9)
Mixed effect models (10)
Survival analysis (11)

Communicate results with quarto

Past

text in Word, code in R
copy figures & tables, paste into Word
find mistakes - go to step 1.

Future

Code and text interwoven
Reproducible, dynamic documents

Data death

Natural degradation in information content associated with data and metadata

Empirical evidence of data loss

Probability of data survival vs paper age. Vines et al 2014

What would happen if you lost all of your research data?

I was focussed on creating high resolution, 3D time lapse videos of developing crustacean embryos, so all of my work was digital-based. When I lost my laptop and backups, I lost 400GB of data and close to four years of work. As a direct result I ended up getting an MPhil rather than the PhD I’d been working towards. I was hoping to have an illustrious career in science and for a time it seemed like everything would be stopped in its tracks.