• There’s always uncertainty in the data, e.g. caused by
    • Measurement errors
    • Too short a window into long-term study
  • It’s important to control for as many interfering variables as possible
    • Example: if you want to measure whether kids pick up the toy that has been shown to be friendly and helpful, you have to make sure that decision isn’t influenced by its shape, colour or position
  • Not all variables can be controlled, there is always randomness
    • We have to apply a probability model to find out how random the choices actually are
    • Question: Would the results be the same if the choices had been made purely by random chance?
    • Example: Say there are 16 babies and 14 of them picked the helper toy. If they had a choice between two toys, the probability for random choice would be 50/50. Getting 14/16 is much more unlikely. This probability is called the p-value.

Reference: Statistical Thinking