Allen Wilcox

How is it that NIH just awarded $170 million to develop “personalized nutrition regimens” when 10% of Americans worry about getting enough to eat? nih.gov/news-events/ne…

The window between learning about a pregnancy from a missed period and fetal cardiac activity is only a few **days** For >20% of people with irregular periods, window may not exist at all. New @PNASNews study with Lindsay Cannon and Allen Wilcox:

Have you been using SGA to represent high-risk infants? Me too. It turns out that SGA doesn’t work as well as we thought. links.springernature.com/f/a/sKLkGI00xX… #epitwitter @societyforepi @sper_news @ser_spc @isee_global @PPE_journal

(and just in case you didn’t know, Trump HAS been vaccinated.) thehill.com/homenews/admin…

I have this idea for a billboard for a county of my state with one of the highest % of Trump voters and the lowest vaccination rates. Would it help?

@EricTopol There is no basis in Qatar paper for saying the two vaccines differ in preventing serious disease or death. Those events were so rare that the confidence intervals around the estimates were both HUGE. Jury is still out.

New and important data on Delta breakthrough infections, large Qatar experience and follow-up of > 1.2 million vaccinees medrxiv.org/content/10.110… 2-doses vaccine effectiveness vs infections Pfizer 53.5% Moderna 84.8% Either 57.2% (replicates gap of recent Mayo Clinic study)

Today’s NYTimes has a long article about Joe Rogan, and mentions one of his guests, Shanna Swan “a reproductive epidemiologist.” Shanna, you’re famous! We’re all famous! @sper_news @DrShannaSwan

@fracardo @societyforepi @SPER_News @PPE_Journal That is fantastic! (Sorry, I don’t check Twitter so often.) How did you ever track that down? Can you provide full ref? Write to me at allen.on.sabbatical@gmail.com. I’ll see your reply faster that way... Many thanks.

@wilcoxEPID @societyforepi @SPER_News @PPE_Journal I found a german source from Johann Samuel Hersch from 1849 defining preterm (=Frühgeburt / partus praematurus ) as birth between 29. and 37. week books.google.at/books/content?…

When did doctors first define preterm as <37 weeks? Ancil (1964) claimed a French physician, suggested it in 1902, so I had article translated. Lots of grim history on infant mort, but nothing about <37 weeks. Still searching! #epitwitter @societyforepi @sper_news @PPE_journal

You may never have listened to Dvorak’s Slavonic Dances as Dvorak meant them to be heard - as piano duos. This in memory of my friend Frank, who died yesterday. We spent many a Sunday afternoon butchering and relishing these wonderful pieces. youtu.be/o16Sa1Z-7Dg via @YouTube

I can’t get in either. Help! #SER2020

Will it be an SER meeting without the hugs? Tweets and likes will have to do for now. #SER2020

via @NYTOpinionb. Has somebody beat me to this? It’s worth watching twice. nytimes.com/2020/11/25/opi…

Our unmarried son in California is spending Thanksgiving alone. Our daughter in New Jersey is shipping him a homemade pecan pie. My wife is sending him the cranberry sauce he loves. Family celebrations come in many forms.

Obama’s election was a triumph, but nothing like the euphoria of being rescued from a burning building.

Less power = more chance of a false positive. Like when we forage in subgroups. Subgroups reduce sample size, reduce power, and increase risk of false positive - even as results may achieve “statistical significance.”

Under the (extremely) constrained conditions of this thought experiment, 4.0 is the most likely to be false and 1.3 is the least likely. Equal P values do NOT mean results have an equal chance of being false. Power also matters.

Thanks to everyone who responded to poll. To recap, “Which RR is most likely false positive? Assume no bias, all P values=0.04, and ALL else held constant.” Responses were “4.0” 38%, “1.3” 19%, “both equal” 43%. #epitwitter @societyforepi @epidemilogyLWW @iseeglobal

@ken_rothman Thanks.

@wilcoxEPID I'm not sure yet where you're going with this, but whatever the objective, I think it would best to visualize the two P-value functions. Here you go.

Help me out with something. Say you have two findings, RR=4.0 and RR=1.3, and both have P=0.04. Assuming there’s no bias and all else is held constant, which would you highlight in abstract? #epitwitter @societyforepi @sper_news @isee_global @PPE_journal @EpidemiologyLWW

@paulpharoah Yes, equal priors.

@wilcoxEPID By all else constant do you mean equal prior?