PHIGHT COVID RESEARCH PROJECT MSP Team: Cheyenne Ehman, Shirley Luo,
Zi Yang, Ziyan Zhu

Seema Lakdawala
Valerie Ventura

Yixuan:

Objectives change:

* How do statewide school poliies affect transmission at the county level
  (Ohio data).  Are children an important factor?

* How do we measure xmit
* Does teach meth affect case/death rates
* What covariates matter?

Zi:

Data sets:

2 data sets:
  - ohio county level cases 36K obs , 17 vars
  - ohio k12 data (demographicds and policies)

Ziyan:

data cleaning and wrangling

omit some variables
manually correct wrong and NA; some rule for dropping when too much
  missing (aggregating to county level I think)                      sl 10
Bracketed data to reopen data and 10 days before xmas break
  deaths increase per population

Cheyenne:

Methods

EDA to find relevant covariates
ANOVA (Dunnet, Sheffe tests, etc for multiple comparisons)
Time series



Ziyan:

Bracketing helps focus and gets significant results for

Yixuan: roadblocks

many important covariates inexact at best

sl 17 online only has a mask vs no mask difference
(is this a policy variation?)

QA partic



--------------------------


HCI -- Skill prerequisites MSP Team: Smeet Poladia, Kelsie Lu,
Elaine Xu


Elaine

Zhou - the data

(lots of verbal description but that's where they are)

describing simple conditional prob estimation as EDA for 0/1 response

EDA for assistance score (need to log the assistance score and/or omit
wierd super extreme outliers)

assistance scores def need to be logged

Smeet - methods

gaussian graphical models

look at clusters, try to understand prereq relats from the ggms'

Elaine - learning curves (prelim illustr)


Smeet - next steps

improve asssistance score to a "mor robust metric"

working on prereq relats from associational data

more predictors for learning curves/afm model

QA

Elaine
Zhou







IFDA Talent PROJECT MSP Team: Lanyi Xu, Malik Khan, Xiaofan Zhu, Yanxi
Zhou, Echo Luan

Malik: introduction

narrowed focus from recruitment + retention to just recruitment

looking at data and WIP for Apr 9
  - sl 2 how much data are you waiting for

BLS 2019 limitation
2 of 5 clients have given data; waiting on other 3
  - sl 3 what they got from each of the two companies

qualitative responses from structured interviews

less model building and more feeding bacl what clients know and don't
know; recommendations for furthet data collection and anasylsis

Xiaofan - deeper into public data

BLS:

  good EDA of BLS data but now connect to IFDA questions
  sl 7 left - purple most relevant?

Yanxi:

client data:

sl8 really interesting that most terminations are for new hires!!!
(is this a recruitment or retention thing? or both?)

Lanyi: 

turnover rate = # turnover / average head count  (2018 and 2019 data)

sl 10 what is "turnover"

intersting regional effects - maybe also some time series effects?

Echo:

hourly wages

sl 12 can we compare these wagse with industry in general and/or with
employers outside IFDA (e.g. amazon)

sl 13 why is termination so mch hiher for semi delivery and warehouse
shipping?

i ddin't undersand sl 14

Malik:

part of our goal is to help clients know what data to collect and
complete

next steps:

recommendation to clients: normalize data between clients!
(same data from each client)