and the datacamp grind resumes!

here are my notes on the first 3 chapters of the datacamp course i did today: experimental design in python:

chapter 1: experimental design preliminaries

• setting up experiments
  • experimental design - a process in which we carry out research in an objective and controlled fashion to ensure we can make specific conclusions in reference to a hypothesis
  • when noting conclusions, use precise and quantified language
  • subjects - what we are experimenting on
  • treatment - change given to a group
  • control - group not given any change
• experimental data setup
  • problems with randomisation
    • uneven issue - diff no. subjects in each group
      • solution - block randomisation
    • covariate - high variability in some covariates - group imbalance
      • solution - stratified randomisation
• normal data
  • drawn from the normal distribution
  • an underlying assumption for many statistical tests (parametric tests)

chapter 2: experimental design techniques

• factorial designs: principles and applications
  • factorial designs allow simultaneous examination of multiple variables
  • this kind of setup tests every possible combination of factor levels, not only measuring the direct effects of each factor but also the interactions between them
• randomised block design: controlling variance
  • reduces variance by grouping similar units, allowing all blocks to receive all treatments
• covariate adjustment in experimental design
  • covariates - variables that are related to the outcome variable and can influence its analysis
  • can help in reducing confounding

chapter 3: analysing experimental data: statistical tests and power

• choosing the right statistical test
  • independent samples t-test - used to compare means of 2 groups, assuming a normal distribution and equal variances
  • one-way anova test - used to compare means across multiple groups, assuming equal variances among groups
  • chi-square test of association - used to test relationships between categorical variables
• post-hoc analysis following anova
  • used after significant anova results to explore pairwise group differences
  • 2 key methods: tukey’s hsd, and bonferroni correction