Today’s excursion into pithy commentary:
- National Novel Writing Month has commenced. I’m stretching my legs a bit to write a sci-fi story. My goal is to have the science be accurate but in the background; I’m really aiming for a commentary on human social evolution that just happens to bet set forward in a non-archetypal future. NaNo requires 50k words. I’m aiming for 90k, because I’d like to brush up the first draft and possibly shop for an agent or publisher. High goals, but hey. Defecate or get off the pot; I’ve been meaning to write a novel for years so why not now?
- To that end, I’m not only participating in local write-ins this November, but I’m also hosting one. On Saturdays at the food court at 35 Michigan. We had the first one yesterday and a baker’s dozen attended. It was great. The Starbucks baristas were friendly, too.
- So just about every female I’ve seen under the age of 30 seems to be wearing skin-tight clothes (including, most frequently, black leggings) paired with boots that frequently rise above mid-calf. Fashion is one thing; lemming-like wardrobes is another. At least the cellulite brigade hasn’t hopped on the bandwagon yet.
- Many of the regular patrons at my local cigar shop are Freemasons. They’re a cheerful, civil lot. Apparently they’re barred from asking people to join up, but they’ve hinted that perhaps I could ask questions about joining up. Alas, the Catholic Church still considers Freemasonry a grave sin. I chatted up another regular about it — a local priest, as it happens — who was quite happy to share his (utterly orthodox) knowledge of the Vatican’s perspective, with an added challenge to “return the favor” by encouraging the masons to consider membership instead in the One, Holy, Catholic and Apostolic Church. Hmm.
- I refreshed my HP TouchPad to the latest release of CM9. Much improved over the version I had installed earlier this year. I also downloaded a desktop-sharing app that works flawlessly with my Win8 PC at home. Meaning, I can keep working on my novel in Scrivener on my tablet without any loss of data or continuity.
- Halloween was somewhat underwhelming. My office door was decorated on my behalf. I ended up doing last-minute NaNo planning with Brittany, Steve and PJ at Wealthy Street Bakery. Very helpful.
- The election looms. I’m planning on watching the returns with Tony in (of all places!) Southfield. He needs to fly to California early the next morning from DTW, so we’re going to watch early returns at Churchills’ cigar shop there and enjoy a fine dram of Scotch or two. I’m cautiously optimistic that Romney will eke out a win, and I may be working Tuesday morning as a poll challenger at one of the most heavily Democratic precincts in West Michigan. Hmm.
- Speaking of the election, there’s been a lot of background noise about Nate Silver, the NYT blogger/prognosticator who’s been consistently “predicting” an Obama win. The whole situation annoys me. Look, as a full-fledged member of the American Statistical Association, I can say for certainty I know what Silver’s doing — he’s assessing the probability of a binary outcome, based on various undisclosed polls as inputs into his model. That’s fine. As a full-fledged member of the Society of Professional Journalists, I can say for certainty that if journalists could do stats American reporting would be of uniformly higher quality. That said, the fundamental problem with Silver’s analysis is that he’s basically got a garbage-in/garbage-out thing going. I don’t question what I understand his methodology to be; I do question the radical differences in polling that feeds his model. Polling in this cycle is all over the map. Throw crap in, you’ll get crap out, no matter how carefully you run your probability estimates. So a pox on everyone’s houses.
- Oh, and on top of it, American politics isn’t accustomed to a binary probability estimate for presidential elections. So moving in that direction, given that the inputs are more volatile than people give them credit for, seems like a misapplication of models. But hey, if Silver wants his 15 minutes of fame, he’s certainly earned it.
- I swapped out the stereo on my Jimmy a few weeks ago. The operation took longer than it should have — I needed to buy a wire harness — but I managed to get the job done without damaging anything. Yay, me.
- Tony I went on a brief casino trip a few weeks ago to Harrah’s Joliet and met Mark and Keren of the 360 Vegas podcast. An uproariously good time was had by all.
- Life at the hospital continues to pay lip service to the Chinese maxim about interesting times. I’m now officially a business analytics analyst in the Information Services team. The transition continues to unfold, so stay tuned.
November, already. Ugh. But hey — the holiday season’s a-comin’. Be glad, and rejoice.
4 comments
You’ve left me no choice but to defend Nate Silver.
First, his predicted electoral college map was what actually occurred. His model worked. Whereas, you,a full-fledged member of the American Statistical Association, predicted that Romney would eke out a victory.
Second, he’s not just a blogger/prognosticator at the New York Times. He’s earned a living as a statistician, and he holds an econ degree from the University of Chicago. He has the credentials.
Third,you say that he’s using “various undisclosed polls as inputs into his model.” Did you read his blog? Those polls aren’t undisclosed! They’re all public. He just weights them based on their past reliability. I’d expect a a full-fledged member of the Society of Professional Journalists to get his facts right.
Finally, President Obama. Deal with it.
Do you actually read anything I write or just spew bile for the fun of it? I didn’t challenge or question Silver’s model; I questioned the premise about how they get weighed.* I also didn’t predict a Romney win, I said I was cautiously optimistic about it — a qualitiative, not quantitative, assessment.
The funny thing is, Public Policy Polling — which also came very close to dead-on this cycle — recently acknowledged that their samples were weighted heavier +D than other people because they just “felt” that more Dems would come out than others were expecting.
Therein lies the problem (or, the “garbage in/garbage out” component) of this whole thing: The formulas work — I never suggested to the contrary — but the decision about what to put into those formulas is still being made by consultants, not statisticians. As I understand it, the Romney camp did something similar to what Silver and what Obama’s data team did, but they used a different set of assumptions about the ratio and strength of who would show up at the polls. Those assumptions shape the way that polls are designed, and hence the quality of data upon which simulations may be conducted.
*BTW, Silver absolutely has not released his internal weights and such, contrary to your assertion.
Do this:
1. Go to http://fivethirtyeight.blogs.nytimes.com/
2. On the right side of the page, scroll down to the section titled:
“State-by-State Projections.”
3. You’ll see for different states the polls that went into his model, and a “538 Wt” for each one. So, he does disclose the weights he gives polls.
Your first assertion was that the polls were undisclosed, then you said that while they were disclosed, he did not release his internal weights. When are you going to start dealing with facts?
Good heavens. Methinks you’re trolling again.
I’m quite capable of navigating FiveThirtyEight. I’ve been reading the RSS feeds for years. As I’ve said before — several times — I’m not questioning his basic methodology.
Silver hasn’t released his weights. Sweetie, a multiplier isn’t an algorithm. I’m glad you think you have a monopoly on facts, but you simply cannot point to Silver’s weighting scheme because HE HASN’T RELEASED IT.
A summary multiplier isn’t an internal weight, no matter how it’s labeled on a website. I can understand why you might be confused.