In-person
session 2
January 18, 2024
PMAP 8521: Program evaluation
Andrew Young School of Policy Studies
Plan for today
Plan for today
Files, folders, and projects
Plan for today
Files, folders, and projects
Transforming data with {dplyr}
Plan for today
Files, folders, and projects
Transforming data with {dplyr}
Regression, p-values, and null worlds
Files, folders,
and projects
Why so much content
these first two weeks?
Everything is a little scattered at the beginning because you're all coming from different backgrounds and skill levels. Some have been using R for months+years, some haven't. Some have been doing regression and stats for years, some haven't. So these first two weeks involve throwing a bunch of stuff at you and seeing what you'll take and what you need.
In the future, it'll be a lot more consolidated: content on the content page, assignment instructions on the assignments page, example walk through on the examples page. That's it. Content, assignment, example, over and over again.
How much should I be reading?
File paths, working directories,
and RStudio projects
- https://www.theverge.com/22684730/students-file-folder-directory-structure-education-gen-z
- Working directories and RStudio projects
.zip files
The hyperliterality of computers
Warnings and messages
Quarto tips
- Visual editor
- Citations - https://zbib.org + https://zotero.org
- Markdown metadata and outputs, figure sizes
Transforming
data with {dplyr}
Regression with R
Regression equations
And is the intercept ever useful,
or should we always ignore it?
Why use two steps to create a regression in R?
(i.e. assigning it to an object with <-?)
Why use tidy()
from the broom package?
Show model with lm(); show t-test with t.test(); show both through tidy()
Use marginaleffects
How was the 0.05 significance
threshold determined?
Could we say something is significant
if p > 0.05, but just note that it is at
a higher p-value?
Or does it have to fall under 0.05?
Why all this convoluted
logic of null worlds?
Do we care about the actual coefficients
or just whether or not they're significant?
How does significance relate to causation?
If we can't use statistics to assert causation
how are we going to use this information
in program evaluation?
What counts as a "good" R²?
R time!