9999

Sunday, December 14th 7:45pm Dear Diary, This is unbelievable. I am so outraged, I can’t hardly believe this hasn’t been made more of an issue. Newly elected commanders-in-chief and their families are allowed to stay at the Blair House five days prior to inauguration. We had just asked if we could would be allowed to move in a few days earlier so that Sasha and Malia could start school in DC immediately after the holiday break. But since Bush is insisting that his pre-scheduled events and guests take precedence over us, we will not be allowed to move in early. Basically, we’re not allowed to stay in the guest house because Bush wants to throw parties for his friends, and they don’t want to do it with the First Family living in the house. I’m curious to find out which guests outranked the incoming President. I am so upset over this. This would never have happened if John McCain had been elected, and he and Cindy were moving into the White House. In that case, the Bushes would have greeted the McCains with open arms, and a bottle of champagne, I’m sure. Believe me, it is very difficult to find a house or apartment suitable for an incoming president in the DC area. It’s not like we’re asking for a mansion or penthouse apartment. We just want a comfortable home where we can live for a few weeks. But all landowners want someone who will live there for a long period of time, not a month. And this is all because we want our daughters to start school on January 5th. If anything, these people should have sympathy for a family prioritizing the education of their children. But per usual, parties are more important to Bush than education. Yours always, Michelle

If I have a pending refill with erxonline who also uses Pharmanet and I'm not due for a refill for another couple weeks, will Pharmanet fill a new prescription for the same meds through buymeds.com? In other words, If I place an order now with buymeds but have an existing order with erxonline and they both use the same pharmacy, is there going to be a problem getting meds??? Any info appreciated.

I own a 1999 Ford Contour. Six month after purchasing the car new I began having air conditioning problems. The refrigerant was replaced forst, then the evaporator then the dealer gave up. The problem that exists now is the dash vents (ac system) produce air which has a 20 degree temperature variation acrodd the dash vents.With an outside temp of 83 degrees the right side vent might deliver 53 degree cooled air while the left side vent delivers 73 degree air. The result is a cabin temperature of 88 degrees. The dealer noted on the last work order that there was a design defect. I don't believe that. The car is still under warranty.

As you know, jemanji is not into quibbling. But this convention on "Sample Size" is both (A) madly out of control and (B) seriously harmful to our analytical thinking. Probably my least-favorite expression these days is "I'm dumber for having [listened to the Hannah Montana theme,]" or fill in the blank. The visual is obvious: I'm so thunderously intelligent that even coming into contact with average people lessens me. Ouch.  I think we could all do without that not-too-subtle implication. But here's a case where, after you read a sabermetric analysis that revolves around the idea of "insufficient sample size," you literally ARE dumber for having read it. :- ) Literally. When you buy into the pseudo-application of Sample Size, your thinking is less effective afterward. .

=== Giving Up the Game, Dept. === The story goes about two Nazi spies who were captured crossing the border because they told the border guard they were 'out of petrol.' In this country they called it gasoline, not petrol ... the SS guys gave up the game by using such odd grammar that it couldn't have come from a citizen of that country. Likewise, the single phrase most popular with sabermetricians, "sample size," is used by us in ways that real statisticians would never use it. . === Real Men and Real Samples Dept. === You'll hear a guy talk about Godzilla's rookie .435 SLG as "a large sample" of NPB-to-MLB performance. (Or at least a large "sample" of Godzilla's own performance.) Then they'll create formulas for predicting Fukudome's stats, and these formulas will calculate to the 1/1000th point of precision... Hey, never mind those 700 PA's; you'll hear people say that Jorge Campillo's 10 innings this year are "a small sample" of his performance, so maybe we should be a *little* careful. After all, it's a small sample, not a large one. A real "sample" is drawn from a scattered set of points across an entire data universe. Not from one small sector of the data universe. .............................. If Boeing wants the FAA to approve a Quality Assurance manual for its 787 -- based on its Statistical QA processes -- they'd better not even dream of "sampling" the wing brackets by taking the .1% that were produced the first day. Can you see why? A "sample" of the year-2007 wing brackets is taken some from January, some from March, some from Tuesday, some from Friday .... you see the point? (I haven't seen the specific QA sampling requirements for fasteners on the -87, but you understand.) A "sample" of Derek Jeter's performance is not the 75 OPS+ he put up as a rookie. Jeter's 1995 performance was not "a small sample." It wasn't a sample at all. It was first returns. The difference is fundamental. Godzilla's rookie .435 SLG was not a "sample" of his performance. A "sample" of his performance -- to a Boeing QA statistician -- would be a game in May 2003, and then a game in June 2003, and a doubleheader in 2004, etc etc. A "sample" of Roger Clemens' performance would be drawn from 39 different ML seasons.... ............................... In fact, if you want to declare something a "sample," you need to meet two conditions, in the business world: 1. It has to be a large enough "sample" that you can use probability theory to prove it's reliable. (This is almost never done by sabermetricians.) 2. That "sample" must be shown to be representative of the entire dataset -- that it wasn't taken from one day when your managers were at the factory, that it wasn't taken from Tuesdays when your workers are at their best, that it wasn't taken from your best factory, etc etc etc. . === Clint Eastwood 'A Man's Got To Know His Limitations' Dept. === It's not that we have "a small sample" of Ryan Rowland-Smith's performance. We don't have a sample at all. ........................... When BP used the "sample" of Godzilla's rookie year to overhaul its NPB-MLB conversions -- which it did -- it wasn't working with a "sample" of Godzilla's performance. (After BP drastically reduced its NPB conversions, Godzilla's SLG promptly went from .435 to .500 the next year, and then stayed there.) No offense to them; I like BP. But they would tell you that their 2004 conversions were "based on the best information at the time." No, because they weren't working with the best information. They were working with data that had been interpreted in a way that was fundamentally unwise. They thought they had a sample. What they actually had was first returns from a moving target. === Give A Guy A Chance, Dept. === This isn't a quibble. Baseball folks -- Good Ole Boys, saberdweebs and us fans alike -- already put way, way, WAY too much emphasis on first returns. No kiddin', if Greg Halman comes up and hits four home runs in his first week, we'll all have him in the HOF. If he comes up and fans 20 times in 30 AB's, we'll all want him traded. Literally. Matt Tuiasosopo had a wretched AA debut in 2006, and mathemeticians concluded that Tui was done, based on those 216 AA AB's. Why? Because "no star ML'er had ever struggled in exactly that way" and because we had a "reasonable sample" of 216 AB's. Now it's clear that Tui is back on the map. We had fallen into a trap: we believed that Tui's results at AA at 21 were representative of his career as a baseball player. No ballplayer's single season represents him from a predictive point of view. No 100 games are going to sample his career. The 100 games aren't random enough. We already put far too much emphasis on any ballplayer's early failure, without implying that it has been statistically proven that Tui or JLo or Godzilla are dog doo. :- ) Most saberdudes really do not grasp the idea of getting a varied and therefore representative "sample." . === Like What, Dept. === Brandon Morrow, early on, walked 47 men in 61 innings. Dr. D sagely and melodramatically cautioned that this is not a 60-inning "sample" of Morrow's career performance. When Morrow gets into rhythm as a SP, and especially when he figures out the happy medium on muscling the ball for an extra 2 mph, those BB will mysteriously vanish. By 2011, nobody will be talking about 47 walks in 61 innings as a "sample" of Morrow's ability. Ryan Rowland-Smith, when he first came up, put up peripherals that would make Cole Hamels proud. But we didn't have a 38-inning "sample" of RRS' outcomes. (if we had, the Mariners would have HAD a Cole Hamels on their hands.) We had a half-year of RRS sneaking up on people in blowouts, and with a good number of LH batters into the bargain. The Hamels-like results aren't coming from a "small sample." They were coming from no sample at all. ................ Miguel Batista? Now, there's a guy where you could take any one game from his entire career, and it would have turned out to have been a "sample." ;- ) Miguel is one of the most reliable and predictable (average-solid) vet SP's I have seen. Good to have ONE of those around. It TURNED OUT that a particular slice of Batista's performances would have been a predictive "sample." Like if you'd "sampled" the first four Tuesdays of 2008, and found they all had 24 hours, and it turned out your sample worked. Jose Lopez? His results are not a "sample," not to a real statistician, because his arc is going up. See if the FAA will certify your airplane based on that kind of sample. His results are a "sample" in one limited sense: he's not going to achieve any .500 SLG's until he jumps plateaus. We know that much. ................... Friends don't let friends use the phrase "sample size," not when that sample is drawn from a single ballplayer's results. Now, if you want to use it in a leaguewide study, that might be a different subject. We get three strikes at once when we use 'sample size' the way we do. 1) The way we use it is mathematically incorrect. 2) The way we use it leads us to wrong conclusions. 3) The way we use it leads us to be overconfident as to the quality of the information we have. Except for that, we should keep on using it the way we always have :- ) Enjoy, jemanji ................... image:  http://staticpixel.com/images/p1250044_ball_bearings_bw.jpg