Specific things to mention

[ivelasq]

[Workshop 2 Chat] best practice to do variable manipulation (mutate) before or after creating design object. My answer is either way is fine
[Workshop 3 Chat] is using survey_glm the same as using glm with weights? Does survey_glm do anything behind the scenes that is different?
-survey_glm accounts for the entire design structure (cluster, strata, etc)

• point (coefficient) estimates would be the same but not variance and thus not p-values
  ^ thank you!
  Couple resources on that:
  (Blog post on the statistics of survey regression) https://www.practicalsignificance.com/posts/survey-regression-what-we-estimate/
  (Lumley’s overview article with statistical details) https://projecteuclid.org/journals/statistical-science/volume-32/issue-2/Fitting-Regression-Models-to-Survey-Data/10.1214/16-STS605.full

Why don't calculate p after N? - variance calculation reasons (SZ)
What is a design effect? - I answered this and certainly will include in book (SZ)
More clarity on conditional vs joint ps (SZ)
Using broom::tidy with svyglm - yes it can be done (SZ)
Difference between two types of filter (before & after group_by and filter) (SZ)
Was asked when to use subset vs filter (RJP)
I was asked when to use R Markdown and when to use R scripts (SZ)
Was told to include why we use survey analysis vs SRS basic analysis in the book (RJP)
Asked to provide more detail on the data like what the clusters/strata are if using them (RJP)
Something that came up was the filter. I think this needs to be presented earlier since we never really talk about that in the categorical section. There is detail to this in the continuous section, but I had to explain that here. All initial exercises in categorical use filter, so we probably need to ensure that we talk about that earlier (RJP)
Maybe also show the different levels of the variables, or walk through one specific example so people understand what to filter on (RJP)
Talked a little about the forcats package as well. This is really helpful for survey data so might be good to include a segment on this when doing derived variables/mutating of data (RJP)

Workshop 2-Good questions:
Difference between survey_prop and survey_mean (from someone who wasn’t in the first workshop)
Meaning of standard error in survey_ratio and more examples of how survey_ratio works
Best way of finding variables from survey object
Difference between glm and survey_glm
How to subset into test/train (open issue on GitHub)