Class 10 — Simulating Data
Part II ended with your pilot study written up start to finish. Today Part III begins with the tool your final project’s data will come from: simulation. Quiz 2 (covering classes 6–9) opens the class. Then you’ll build a machine that generates what a whole different sample of participants running your study might have looked like: simulate condition means for N hypothetical participants with rnorm(), using the pooled between-subject SD computed from the whole class’s HW3 results, repeat that a thousand times with replicate() to build a distribution of study-level effects, and compare that distribution to your own real HW3 result, thus building informal intuition about power and sample size.
Reading due: Data Skills ch. 10 · Due today: — (HW5 was due Mon 11/9, in the no-class week, so nothing is ever due the same day as a quiz) · Quiz 2 at the start of class
What you’ll be able to do
- Recall and apply RStudio/tidyverse fundamentals (including the Git pane), dplyr wrangling, ggplot2 visualization, and reproducible-reporting concepts from Classes 6–9 (Quiz 2).
- Explain, in your own words, what a simulation is and why psychologists build fake data on purpose.
- Explain why within-subject variability (how noisy one person is, trial to trial) and between-subject variability (how different people’s averages are from each other) are genuinely different things.
- Write an R script that draws random samples from specified distributions (
rnorm()) to simulate N hypothetical participants’ condition means for a 1-factor/2-level design. - Run that simulation many times (via
replicate()) to generate a distribution of study-level effects, and visualize it with ggplot2. - Compare that simulated distribution to your own real HW3 summary result, and describe informally what that comparison suggests about power and sample size.
