#rstats backbone infrastructure of library(ipumsr)
@ipums ipumsr relies on the @DDIAlliance DDI code book metadata format for approximately everything
Some internal structure of the DDI codebook objects in ipumsr
Variable names and labels and value labels are available
Your daily reminder that #rstats factor variables suck big time compared to Stata and SAS
I am not crying, you are crying
Helper functions: replace missing values. These suck in #rstats compared to Stata and SAS, too: in the latter, you can have extended missing values .a:.z, and they all can be labelled with the reasons it is missing (not in pop, refused, not applicable, etc.)
Another helper: recode and label new values. Very helpful for hierarchical classification codes where you can divide by 10 or 100 to obtain the higher level / fewer digits code. Labels are copied from the lowest numbered original category - would need extra work down the line
A more generic function for that is relabel.
Now let’s talk about @ipums@nhgis geographic data - tied to library(sf)
There are helper functions for the geo data management
Coming soon - @popdatatech API - ask to beta test. Sharing extract definitions is important for reproducibility!
By how much do response rate differ between the different population groups? Today, we will explore this with the @BRFSS data (cdc.gov/brfss/annual_d…) 🧵👇 1/11
The data set makes it possible to study these differentials as it is one of the relatively rare data sets with both the design weights and the calibrated weights... in our case, _WT2RAKE and _LLCPTWT. 2/11
(It's not me who likes to YELL, it is CDC. If you don't like the variable names, janitor them the way do like) 3/11
Quality of the #2020Census data -- a session on race and ethnicity coding.
Census follows the OMB 1997 standards, does not set the standards: 2 categories of ethnicity, 5 categories for race (AIAN, Asian, Black/AA, Native American/Pacific Islander, White) + "some other race"
The race and ethnicity questions help understand how people self-identify, so research into these is necessary to understand how the U.S. population evolves (more multiracial, more diverse than measured in the past)
There were some proposals to start offering "Middle Eastern / North African" (MENA), but they did not make it to the #2020Census.
#JSM2021 panel led by @minebocek on upskilling for a statistician -- how to learn??
@minebocek#JSM2021@hglanz no shortage of stuff to learn. First identify what you don't know -- that comes from modern media (blogs, twitter, podcasts; groups, communities -- @RLadiesGlobal or local chapters; professional organizations -- @amstatnews ).
#JSM2021 an exceptionally rare case of ACTUAL out of sample prediction in #MachineLearning#ML#AI: two rounds of the same health data collection by @CDCgov
@CDCgov Yulei He @cdcgov#JSM2021 RANDS 1 (fall 2015) + 2 (spring 2016): Build models on RANDS1 and compare predictions for RANDS2
#JSM2021 Yulei He R-square about 30%; random forests and grad boosting reduce the prediction error by about 4%, shrinking towards the mean; standard errors are way to small (-50% than should be)
1. when will the survey statisticians in the U.S. move from weird variance estimation methods (grouped jackknife) to simple and straightforward (bootstrap)
and
2. when will they move from weird imputation methods with limited dimensionality and limited ability to assess the implicit model fit (hotdeck) to those where you explicitly model and understand which variables matter for this particular outcome (ICE)?
Oh and somebody reminded me of
3. when will we move from PROC STEPWISE to lasso as the rest of statistics world has
#JSM2021@jameswagner254 Using Machine Learning and Statistical Models to Predict Survey Costs -- presentation on the several attempts to integrate cost models into responsive design systems
#JSM2021@jameswagner254 Responsive designs operate on indicators of errors and costs. Error indicators: R-indicator, balance indicators, FMI, sensitivity to ignorability assumptions (@bradytwest@Rodjlittle Andridge papers).
Some decisions are made at the sample level (launch new replicate, switch to a new phase of the FU protocol), others at case level (change incentive amount, change mode)