29 August 2017

We Are The World

Yesterday's cautionary tale of dog ownership began thus:
Perhaps the defining characteristic of humans is the drive to divide. We classify just about everything we know, or just think we know, about the world into, generally, two groups. Us and them. Catholics and Protestants.

I didn't know then that Gina Kolata would take up the theme today. Hers is not the first news report I've seen dealing with the trauma of white folk finding out they're not really. And, no surprise, much more of the white supremacist South isn't lily white than elsewhere.
The chances of having African ancestry were highest in the South, and highest of all in South Carolina, where at least 13 percent of those who said they were white had African ancestors.

Boy, that has to burn some, going down. Like chugging grain alcohol. Good on you.

And, of course, there's the other side
Among those who said they were black, genetic ancestry over all was 73.2 percent African, 0.8 percent Native American and 24 percent European. Experts say the large proportion of European DNA found in African-Americans can be traced to before the Civil War, and the rape of enslaved African women.

I wonder if the Klan will admit them. After all, they're more tainted white than whites are tainted black. Kumbaya.

25 August 2017

Dueling Banjos

Once again, into the data fray. This time it's not just a trivial pissing contest. Now, this is a hard data field, not some social science feud over macro policy based on fudgy sampling data. It's all about super-priced heart drugs.
Not surprisingly, each side takes issue with the other's study design and data. They question each other's assumptions about the risk of heart attacks, strokes, and other cardiac events - a key figure that gets plugged into their computer simulations - and they use different thresholds for the proper value of a year's worth of "quality of life."

Well, yeah, there is a bit of social science thumb on the scale. Moreover,
There is another point of divergence in the Fonarow and Kazi studies. When they modeled cost savings from evolocumab, the former factored in productivity - the economic cost of a heart attack victim not going to work, for example. Kazi says his group's study sticks to just the medical costs saved - the hospital stay, the follow-on procedures - by avoiding a cardiac event.

One of my pet peeves with the social science use of data, the majorest, is the continued imputation of "loss of productivity" for all manner of factors. Pharma has been executing "outcomes based pricing" for a while. This is just a fancy way of saying, "my drug saves your life, so you owe me the rest of your life's earnings". I read one "estimate" that the eclipse "cost" $700 million in "lost productivity". As if public and private bureau rats not sitting at their desks for a few minutes has any impact at all. Few do anything productive, ever. A few minutes burning their corneas, on balance, prevented more London Whales from breaching than lost useful output. The one and only case of directly measurable productivity is piece work. Tender reader, you may never have heard of that. Here it is.

18 August 2017

Hi Ho Steverino

As, I suspect, most folks have done, I took it as an article of faith that Bannon's fortune (whatever it may be) was made at Goldman. The fact that he pronounced himself a Leninist never made much sense. Nor, as mentioned in earlier missives, how he or Donald J. Quisling or any President could compel corporations to practice economic nationalism. You have to be a dictator to force corporations to behave against their near-term (aka, the CxO class's) best interest. Do I figured Donald J. Quisling wet dreams of himself as Orange Julius Caesar? But of course. Start a war. Declare martial law. Suspend Congress. Declare himself Dear Leader (he got a bit ahead of the game at that first Cabinet meeting. yuk.).

So, I let my fingers do the walking through the innterTubes, and came across this diatribe from last spring. Sounds much like what one might expect the Oracle of Delphi to say.
In 1985, Bannon was a Navy vet in his early 30s fresh out of Harvard Business School who went to work at Goldman Sachs working in M&A. He lasted there until 1990 when he left as a VP to start his own boutique M&A firm. Now, for the kids and Millennials out there, being a VP at Goldman in 1990 was even more junior than it is now and boutique M&A was the late 80s Wall Street version of a craft cocktail bar; they seemed to pop up everywhere and you definitely knew someone who knew someone who started one.

The stake in the heart:
A born-again blue-collar class warrior like Bannon cannot look back fondly on years of running models 20 hours a day in the pre-internet era only to realized that 33-year-old analysts with no prep school connections had limited futures at Goldman Sachs.

I guess Steve has feet of clay.

17 August 2017

About That CEO Council

Bannon has been gulled into talking plainly in interviews that went live last night (here, and here. They're worth reading in total, but one point is germane here: is there really such a thing as economic nationalism?

Of course not. The CEOs quit just because they want all that globalism. They've spent decades destroying the middle classes, particularly the blue collar segment. They could have partnered with unions in the North to support the blue collar middle class, but instead went South and overseas. Nationalists, economic or otherwise?? In a pig's eye. Which brings us to the real data, finally, on their game. Here's the nut. Autos have been the vanguard of killing labor. The white collar Northern Blue states have been less impacted, in the way explained by Baumol. Whether Watson-ish machines will kill off these folks too remains to be seen. What's the end game, or the look of the world in the limit? The principle problem of rising capital in production is that unit fixed cost rises: you can't layoff a machine, you must still pay for it. So long as output remains steady or rises to the capacity of the capital, you can get away with it. But as capital's share of GDP rises, there's by definition fewer bucks to buy output; aggregate demand declines. A death spiral. Oops.

FRED data shows that GDP going to labor continues to decline. Just today, there's more reporting on the puzzlement within the Fed that inflation just won't come back.
"It would not be desirable," the minutes said, "for the current regulatory framework to be changed in ways that allowed a re-emergence of the types of risky practices that contributed to the crisis."

Good luck with that. Social Darwinism in finance is what the Right Wingnuts dream about all day and night.

The public number, U3, is increasingly viewed with suspicion, just because at under 5% we've always seen labor incomes rise in the past, but not now. There has to be a reason. And that reason is total employment as measured by U6 isn't nearly as pleasant. Wages remain stagnant. Demand remains stagnant. Growth doesn't happen. The rich get richer and the poor attack each other.

15 August 2017

Fake News

Some news really is fake. Here's the headline from my CBS News front page: "The country where Down syndrome is disappearing". A reasonable person might infer from that sentence that some country (Iceland) has somehow found a "cure" for Down. Or at least has found a way to keep it from "spreading" into further generations. Like polio or smallpox. Ya think?

Well, no. Iceland just is more resolute in aborting tainted fetuses than most countries. There is no data that the incidence or prevalence of Down is diminishing in Iceland. There is, as yet, no evidence determining a cause. Down isn't inherited, as recessive characteristics are. It's not race based. It is idiopathic.
Down syndrome is caused by a random error in cell division that results in the presence of an extra copy of chromosome 21.

13 August 2017

Touching Me, Touching You

So, let me get this straight, once again. Clinton lost because 77,900 more shit kickers in shit kicking counties in three states voted for Donald J. Quisling than for her. Explain to me how any Democrat could reach out and touch the votes of such deplorables? Not Dixiecrates, but real Democrats.

10 August 2017

Second Thought for the Day, 10 August 2017

[T]hey've got to draw in their horns and stop their aggression, or we're going to bomb them back into the Stone Age. And we would shove them back into the Stone Age with Air power or Naval power -- not with ground forces.
-- Curtis LeMay/1965

Thought for the Day, 10 August 2017 [update]

As mentioned before in these endeavors, and bears repeating today: the SOP of would be dictators is to instigate war, institute martial law, and claim all governance in the name of protecting the homeland. The shit kickers who bought Donald J. Quisling's white supremacy spiel will buy this transparent ruse. Will the rest of us?

[update]
in answer to a comment in another version of these endeavors, I offered:
Well, another few factoids to consider:

- Donald J. Quisling claims to be second only to Lincoln as President

- the White Power folks consider the Civil War to have been instigated by Lincoln

- Lincoln did institute martial law

Quisling would be doing the right thing.

07 August 2017

Oh My God!!

This missive began in the pre-sleep twilight of last night, wherein the idea of server-centric applications remains ever more true, given the soon to be death of the java plug-in, aka applets. I went to CSC a lifetime ago to do, as it was told to me, "heavy weight database development". Turned out to be VSAM files ported to DB2 on z, run by the existing 40 year old COBOL with revised copybooks and a bit of front-end lipstick. Yikes!! Still COBOL doing RBAR against file data. The products actually ran shitier with the port. Enough so that a few, wise, clients stayed with the VSAM release. Since source code licensing was generally used, such clients took over the modification and extension.

The new and improved version used an applet as the browser front-end, and one of my retread COBOL java colleagues told a prospect "we prefer to do transactions in the client". Heavy weight database development, my eye. A total waste of DB2 fees.

In the olde days of COBOL/CICS/VSAM, the distinction twixt client and datastore was a bit fuzzy, since it's all COBOL and CICS controls the transaction.
The true beauty of the CICS/DB2 attachment is that it's designed to be a two-phase commit process.
(Or, one might say, it's two TPMs in a death match! You have to read that piece, which I found after I wrote my snark; it's a hoot.) But the danger of being hung out by an unsupported infrastructure is minimal. After all, no one got fired for buying IBM.

So, reports re-appeared recently that java 9 will be minus the plug-in. No more applet. When I returned to quant employment, the major aspect was Progress 4GL/database applications. As it turned out, the Progress database at that time was SQL-86 compliant; which means it was just grandfathered into The Club. So, the 4GL was a RBAR language against indexed tables/files. The thing about proprietary languages is that you're betting that the vendor will out survive your business. Progress has. The plug-in and applets haven't.

So, today we find Godzilla SSD release.
The PM5 SAS SSD sets new records by offering capacities of up to 30.72 TB in a 2.5" form factor, and sequential transfer speeds of up to 3350 MB/s (reads) thanks to support for four-port SAS MultiLink. SAS has always supported dual-port drives, with the two ports either used in a failover configuration for high availability or bonded for high throughput.

The end result of all this is that highly (if not massively) parallel processing of relational data on the server is now really quite cheap. Time to treat the browser as an xterm; just another I/O device with all the smarts in the database. The gates to the Emerald City are just ahead. Let's build our applications the way Dr. Codd told us to. OK?

02 August 2017

Farther on Down the Road

The end of the Yellow Brick Road is running R inside the RDBMS. PL/R for Postgres was the first I worked with. The commercial databases offer similar support, but at Big Bucks. Alas. Today we find a new odbc package for R which makes a half-step farther down the Road.
But the real power comes in being able to use high-level functions from the dplyr package and have the data processing run in the database, instead of in the local R session.

If nothing else, this package admits that RDBMS semantics (aka, joins and such) belong there, not in R. Come on guys!! If you can figure out a bastard language like R, SQL is a piece of cake.