Setting expectations and meeting our data

• explain how Bayes is different
• set reasonable priors for your models
• fit, interpret, visualize, and report the three most common models

\(rt \sim Normal(\mu,\sigma)\)

\(\mu \sim Normal(0,3)\) \(\sigma \sim Exponential(1)\)

• feel confident adopting this approach right away
• be ready to fit all models you currently use

You may want to think twice about going Bayesian if you have:

• a simple model
• lots of data
• little prior information

• create a safe space (questions are critical)
• foster comprehension of fundamental aspects
• emphasize creating a reasonable prior model
• touch on all aspects of the Bayesian workflow

• Balancing content: I’ll be checking in and adjusting

• Balancing content: I’ll be checking in and adjusting

• This stuff is hard: a burning sensation in your brain is normal

• One data-set with only native listeners
• Second data-set with native and non-native listeners

rt (reaction time):

• How long from stimulus onset to button press
• recorded in milliseconds
• positive real number

acc (accuracy):

• Whether the response was correct (1) or incorrect (0)
• binary variable

log_Freq:

• unigram frequency (log-transformed)
• continuous variable ranging from 1 to 5

scaled_Freq:

• standardized version of log_Freq

In the course folder (STEP_2023_Bayes) you have a folder called data with the following files:

• LD.Rda and LD.csv
• LD_L2.Rda and LD_L2.csv

You can use the load() function for .Rda files and read.csv() files for .csv files as you’d prefer, I will be using the .Rda files

# Load the LD data-set using the .Rda file
load("data/LD.Rda")

# Using the .csv file
LD <- read.csv("data/LD.csv")

• Load the data (see script S1_E1_data_loading.R)
• Explore the data
• Check in and ask any questions