Comparing randomization inference and t-inference
When technical conditions (see next chapter) hold, the inference from the randomization test and the t-distribution test should give equivalent conclusions. They will not provide the exact same answer because they are based on different methods. But they should give p-values and confidence intervals that are reasonably close.
Cet exercice fait partie du cours
Inference for Linear Regression in R
Instructions
- Calculate the absolute value of the observed slope,
obs_slope. - Add a column to
perm_slopeof the absolute value of the slope in each permuted replicate. The slope column is calledstat. - In the call to
summarize(), calculate the p-value as the proportion of absolute estimates of slope that are at least as extreme as the absolute observed estimate of slope.
Exercice interactif pratique
Essayez cet exercice en complétant cet exemple de code.
# The slope in the observed data and each permutation replicate
obs_slope
perm_slope
# Calculate the absolute value of the observed slope
abs_obs_slope <- ___
# Find the p-value
perm_slope %>%
# Add a column for the absolute value of stat
___ %>%
summarize(
# Calculate prop'n permuted values at least as extreme as observed
p_value = ___
)