Pattern Recognition
 
 [Most Recent Entries] [Calendar View] [Friends]

Below are the 20 most recent journal entries recorded in T. Rev's LiveJournal:

    [ << Previous 20 ]
    Wednesday, May 9th, 2012
    2:53 pm
    ...
     Well, that didn't work.
    Friday, May 4th, 2012
    7:42 pm
    HELLO AMERICA
     EVERYTHING IS UNDER CONTROL. YOU ARE SAFE. THE MOLE PEOPLE ARE COMING. DO NOT BE ALARMED. OH GOD RUN
    Sunday, April 22nd, 2012
    1:19 am
    Oh, Netflix.
    The Man from Nowhere 2010 R 119 minutes
    Former special agent Tae-Sik takes matters into his own hands, rushing to aid a young girl he befriended who was kidnapped by a vicious drug gang. More Info

    Starring:
    Bin Won, Sae-Ron Kim
    Director:
    Jeong-beom Lee

    Based on your interest in: Ponyo, Princess Mononoke and Spirited Away
    Tuesday, April 10th, 2012
    5:14 pm
    Most Scientific Reports are Wrong Postscript: It's Not That Bad
     The Venet et al. paper did get published, after all (though after multiple rejections from oncology journals), and suggests at least two valuable results, one specific and one general.

    First, it gives strong evidence that cancer cells are more prone to expressing genes in general. That seems like a valuable high-level insight. It's not that the previous studies were wrong, precisely, so much as they were missing the point.

    Second, it shows that positive controls and synthetic data sets can be valuable tools for validating research with complicated bioinformatics, where traditional statistical methods can be misleading.

    It also shows that data sets and code should be considered part of published research and should be openly accessible as a matter of course, but I hope that is uncontroversial by now.

    Related discussion over at In the Pipeline: Biomarker Caution.
    4:01 pm
    Most Scientific Reports are Wrong III: That Boom Is the Sound of Scientists Getting Pwnt
     Taken from the study referenced in the previous post:



    ( Unedited chat log of me trying to explain the chart follows. Might clean it up later. )

    TL;DR version:
    • The horizontal axis is the supposed significance of a study. Further to the left is better.

    • Red dots are actual published results.

    • The blue line at the right is the traditional p < 0.05 significance line: a completely random result should only fall to the left of the line one time out of twenty--but keep reading.

    • In each row, the yellow spindle represents a thousand completely random gene signatures, of the same length as the one used in the corresponding red dot, which the authors of the current study tested for significance in the same way using public data.

    • The green bands at the left represent the 'best' 5% of the random gene signatures, i.e. the green band is the best 5% of the yellow spindle.

    • In other words, the real 5% significance cutoff is nowhere near the blue line--it's the green band.

    • The little vertical black ticks are the median random outcomes--any red dot to the right of a black tick is worse than one would expect of a completely random result.

    • A reasonable fraction of the studies made it into the green band (p < 0.05)...but they're analyzing 47 studies, so one might reasonably expect two or three to land in the green band purely at random. (And actually, we should expect more than that, see below.) None landed completely outside the zone of the yellow spindles, which can be taken as a (very rough) proxy for significance at the p < 0.001 level.

    • It's even worse than that, because we're not really looking at 47 studies. We're looking at 47 studies that got published, plus an unknown number of studies that didn't, because their results were not considered significant (they would fall too far to the right on this graph.)
    2:40 pm
    Most Scientific Reports are Wrong II: Noise Looks Like Signal
     Most Random Gene Expression Signatures Are Significantly Associated with Breast Cancer Outcome:
    ..Researchers often suggest that a biological mechanism is relevant to human cancer from the statistical association of a gene expression marker (a signature) of this mechanism, that was discovered in an experimental system, with disease outcome in humans. We examined this argument for breast cancer. Surprisingly, we found that gene expression signatures—unrelated to cancer—of the effect of postprandial laughter, of mice social defeat and of skin fibroblast localization were all significantly associated with breast cancer outcome. We next compared 47 published breast cancer outcome signatures to signatures made of random genes. Twenty-eight of them (60%) were not significantly better outcome predictors than random signatures of identical size and 11 (23%) were worst predictors than the median random signature. More than 90% of random signatures >100 genes were significant outcome predictors...

    ...We demonstrated that in breast cancer any set of 100 genes or more selected at random has a 90% chance to be significantly associated with outcome. Thus, investigators are bound to find an association however whimsical their marker is...
    Emphases mine. Found via a comment at In the Pipeline.
    Monday, April 9th, 2012
    1:53 am
    Saturday, March 31st, 2012
    3:03 pm
    New Scientific Report: Most Scientific Reports are Wrong
    Part way through his project to reproduce promising studies, Begley met for breakfast at a cancer conference with the lead scientist of one of the problematic studies.

    "We went through the paper line by line, figure by figure," said Begley. "I explained that we re-did their experiment 50 times and never got their result. He said they'd done it six times and got this result once, but put it in the paper because it made the best story. It's very disillusioning."
    Link via, and discussion at, In the Pipeline.
    Wednesday, March 28th, 2012
    11:21 am
    THEY PAY COPS TO SHOOT PEOPLE
    In Albuquerque, police have been involved in 23 shootings since January 2010, with 18 ending in fatalities. Critics say those numbers are far too high for a city of 550,000 and link the deaths to a practice of union payments to officers involved in such shootings, calling the union program a "bounty."

    And the police union defends the payments, saying the checks for up to $500 help cover expenses for officers and their families "to find a place to have some privacy and time to decompress outside the Albuquerque area."

    THEY PAY COPS TO SHOOT PEOPLE

    THEY ARE PAYING COPS TO SHOOT PEOPLE
    Monday, March 19th, 2012
    8:15 pm
    SK>JFH kehagt akj
     asjh kjash jkh auheoui jkla

    ajk;egthu
    Friday, March 9th, 2012
    7:50 am
    Friday Mashup
     It is rare for a work of art to make me cackle audibly with delight. This mashup accomplished it twice.

    Daft Punk Vs The Beatles - 'Eleanoround The World' - THE ORIGINAL @daftbeatles MASH UP by DAFT BEATLES
    Wednesday, February 22nd, 2012
    10:54 pm
    Get Happy
     Is it still winter? I can't even tell. I've been sick for weeks, I have to find a home for a kitten, I have another six hundred pages of reading to do, and my right eye is trying to escape its socket. It's time to get HAPPY! Here are two things to be happy about:

    Via It's Got Disco Potential, a perfect slice of electronic pop. DELICIOUS:

    Futurecop! feat. Diana Gen & Starrset - Starworshipper (Lenno Remix) by Lenno

    Via Comics Alliance:



    The prospect of more Phonogram--or any project from Gillen and McKelvie--makes me squeal like a little girl. Or would if I were capable of anything higher-pitched than a Barry White croak right now.
    Tuesday, February 7th, 2012
    12:04 am
    Pokey
     This is Pokey.



    He's the only surviving kitten from Little Mama's recent litter--Little Mama being the sole survivor of Miss Lady's last litter, thus Ratty's younger sister--and is about four months old. Little Mama abandoned him outside the kitchen door a few weeks ago.



    He is very sweet and cuddly, but with three cats already, we can't take him in.

    Wednesday, January 25th, 2012
    9:14 pm
    Is It Time Yet
     For '80s-styled retrofuturism? You know...cyberpunkpunk?
    Thursday, January 12th, 2012
    1:39 am
    Sup
     Hey.
    Friday, January 6th, 2012
    1:16 am
    Hell Yes
     Classic 1985 martial arts/comedy/horror film. I love this movie so much.



    To get subtitles, click the little 'cc' button at the bottom of the screen.
    Sunday, January 1st, 2012
    5:10 pm
    2:16 am
    To [info]twoeleven:
    What about Kowloon?
    Wednesday, December 28th, 2011
    3:58 pm
    Today's Lunch
     Brussels sprouts in manchego cheese sauce with slivered bacon.
    Wednesday, December 21st, 2011
    1:21 am
    WHAT THE HELL
     WHO WRECKED THE INTERFACE

    GOD DAMN
[ << Previous 20 ]
About LiveJournal.com