Powered by glolg
Display Preferences Most Recent Entries Chatterbox Blog Links Site Statistics Category Tags About Me, Myself and Gilbert XML RSS Feed
Saturday, Aug 18, 2012 - 19:07 SGT
Posted By: Gilbert

Some Things

"As long as you work hard and get things done,
you can steal a little, but don't behave like dacoits (bandits)."

- Uttar Pradesh Public Works Minister Shivpal Singh Yadav, in an endearing moment of rare honesty by a politician
(Note: the local solution would be to raise the pay until the problem disappears - same as with indefinite detention, once it's codified, it is automatically proper [though the U.S. did deserve to get pwned on their hypocrisy too])



Some Endeavour


More in-toilet advertising on campus


I do appreciate effort, but what took the cake is not the sticker above, but what happened when I dropped by for a dollar ice-cream at the neighbourhood market - I was about to pay for it, when an insurance/financial services salesman (or are they officially representatives nowadays?) stepped in and offered - practically insisted - to pay for it, if I would fill in a short one-page survey. Foot-in-the-door's upped a level.

But hey, free ice-cream!

Tangentially, I was surprised to be blithely unconcerned when a rat the size of a pair of rolled-up socks darted by my feet when I was in line for an iced Milo. My hamsters should apply for a position on the National Integration Council.


Something Not Adding Up



See, we studied it... but we don't understand it
(Source: xkcd.com)


When a report surfaced that 47 of 53 landmark experiments (some 88.6%) in cancer research could not be replicated, I was not too taken aback, this being a longstanding personal bugbear. However, as stated, this is (still) rarely a matter of overt fraud. The article suggests three main causes:

  • Confirmation bias, or "I need a paper urgently, so this must work!"
  • Poor experimental methodology
  • Luck

These causes do overlap. For instance, a new method may have experiments designed or picked to show that it is superior to existing methods in those respects, which may often be a case of tweaking it until it outperforms on the chosen datasets. In this case, the results are strictly valid, but may not carry over to dissimilar (or even similar but different) datasets, especially without all the minor implementational details, which are seldom reported.

Methodology also has a direct overlap with luck, in that proper methodology minimizes the role of luck. Recalling confidence intervals in beginner statistics, the take-home message is that the more often one repeats an experiment, the less likely that the aggregated outcome is based on fortune (which does also eat away at skill, but that's another matter for next time)

In practice, though, the amount of data/experiments available is limited (e.g. here's the figures, if you're unhappy go build your own particle accelerator), subject to many complicating variables (e.g. when working on cell cultures, is one sure that atmospheric pressure does not matter? Or gravity? Or...), or not even replicable (e.g. human responses in the social sciences), and scientists may have to make do with what they've got. It is then up to them to avoid making bolder claims than they can justify, and also for other scientists to avoid reading too deeply into and generalising too widely from past experimental results.


Something Doesn't Quite Make Sense


No, you tell them
(Source: redcafe.net)


By most measures, fifteen to twenty-four million pounds for probably the best striker in the league for the past year is far from outrageous (Andy Carroll, Stewart Downing, Jordan Henderson, ahem), but don't United need a midfielder, as they have for like, the past decade? Or is it all a sinister ploy to boost shirt sales for both clubs? Are United going to rely on Carrick, a young-and-learnin' Cleverly, a slowly-widening Anderson and the near-octogenarian duo of Giggs-and-Scholes? Stay tuned, and play fantasy football!

Man vs. Ham (now also officially known as Ham vs. Man, due to Esquire Pants' legal intervention) makes its long-awaited return, with Mr. Ham starting with a 50 seed bonus, as previously agreed. Same rules, 100 seeds per week, but in an advance from previous editions, a handy summary page will (should) be maintained by the staunchly neutral Mr. Robo.


Mr. Ham (50 seeds): 90 seeds on Tottenham to get an away win against Newcastle (at 2.50), and ten on Southampton to draw Manchester City (6.50); gonna be my season!

FAKEBERT (0 seeds): Banking on a testy start: 25 each on draws between Arsenal/Sunderland (4.50), West Bromwich Albion/Liverpool (3.45), Newcastle/Tottenham (3.20) and Reading/Stoke (3.30)

[Also: EPL Schedule for the new season from TNP, United matches highlighted. Okay, it's for my own reference]



comments (0) - email - share - print - direct link
trackbacks (1) - trackback url


Next: Some Explaining


Related Posts:
Hamming It Up
Roll Up Roll Up
Pieces From The Past
Heat's On
999-RIP-OFF
Back to top




1 trackback

Trackback by top weight loss pills
top weight loss pills - [bert's blog]


July 19, 2014 - 16:02 SGT     

Copyright © 2006-2024 GLYS. All Rights Reserved.