Page 22 of 22 FirstFirst ... 12202122
Results 211 to 213 of 213

Thread: A little G25-K7 based "calculator" for Europeans only.

  1. #211
    Moderator
    Posts
    5,036
    Sex
    Location
    Normandy
    Ethnicity
    northwesterner
    Y-DNA
    U152>L2>Z367
    mtDNA
    H5a1

    Normandie Netherlands Friesland Finland Orkney
    Huijbregts: I have a personal and shameful experience of those pseudo-scaling operations. When I started to use smartpca, because often of bad choices of populations or of a too small references list, I was always getting "flat" PCAs (eigenvalues too slowly decreasing), and I used to cheat, multiplying the components by a positive decreasing function of the index (most often a geometrical progression with common ratio 1/something). But it was only for personal purpose, as most often that was PCAs with people from my familial clan added to european references from 1000genomes, and I never thought to make from this a canonical operation. In fact it's exactly what the pseudo-scaling does, and nothing else. I agree 100% with your last sentence, and that's basically what I answered to David in one of our debates. A good PCA does not need scaling, and his PCAs ( in particular G25), are amazingly good.
    En North alom, de North venom
    En North fum naiz, en North manom

    (Roman de Rou, Wace, 1160-1170)

  2. The Following 2 Users Say Thank You to anglesqueville For This Useful Post:

     JMcB (11-13-2018),  Ruderico (11-13-2018)

  3. #212
    Registered Users
    Posts
    604
    Sex
    Location
    Netherlands
    Ethnicity
    South-Dutch
    Nationality
    Dutch
    Y-DNA
    I2a2a1b2-CTS1977
    mtDNA
    H13a1a1

    Netherlands Belgium
    Quote Originally Posted by anglesqueville View Post
    Huijbregts: I have a personal and shameful experience of those pseudo-scaling operations. When I started to use smartpca, because often of bad choices of populations or of a too small references list, I was always getting "flat" PCAs (eigenvalues too slowly decreasing), and I used to cheat, multiplying the components by a positive decreasing function of the index (most often a geometrical progression with common ratio 1/something). But it was only for personal purpose, as most often that was PCAs with people from my familial clan added to european references from 1000genomes, and I never thought to make from this a canonical operation. In fact it's exactly what the pseudo-scaling does, and nothing else. I agree 100% with your last sentence, and that's basically what I answered to David in one of our debates. A good PCA does not need scaling, and his PCAs ( in particular G25), are amazingly good.
    I too agree 100% with your last sentence.

  4. The Following 2 Users Say Thank You to Huijbregts For This Useful Post:

     anglesqueville (11-13-2018),  JMcB (11-13-2018)

  5. #213
    Moderator
    Posts
    5,036
    Sex
    Location
    Normandy
    Ethnicity
    northwesterner
    Y-DNA
    U152>L2>Z367
    mtDNA
    H5a1

    Normandie Netherlands Friesland Finland Orkney
    Quote Originally Posted by anglesqueville View Post
    Yes, reason why in smartpca we have the possibility to set the Norm option to "NO". About words, there has always been a part of fuzzyness in the lexicology of statisticians, and even sometimes in their definitions, which is very shocking for "pure" mathematicians. But statisticians are by nature more pragmatic than them. That said, we are on here most often very far from that, as so many guys keep to believe that a PCA is any euclidian graphic in dim 2, and that "Eigenvalue" is the name of an individual with a german father and an english mum . Reason why, actually, a minimal theory doesn't hurt, even if it would be a bad thing to get obsessed by it (which is likely my natural tendency, due to professional deformation).
    I told a bullshit: the option "usenorm" (that I always let, as it is by default) has nothing to see with that. Seemingly with smartpca the variance-covariance matrix is always normalized.

    usenorm: Whether to normalize each SNP by a quantity related to allele freq.
    Default is YES. (When analyzing microsatellite data, should be set to NO.
    See Patterson et al. 2006.)
    En North alom, de North venom
    En North fum naiz, en North manom

    (Roman de Rou, Wace, 1160-1170)

Page 22 of 22 FirstFirst ... 12202122

Similar Threads

  1. Replies: 0
    Last Post: 06-24-2018, 08:50 AM
  2. Replies: 15
    Last Post: 01-30-2017, 09:51 PM
  3. "Indus Valley" and "Kalash" admixtures in Khvalynsk?
    By Tomenable in forum Ancient (aDNA)
    Replies: 12
    Last Post: 08-12-2016, 10:46 PM
  4. Replies: 43
    Last Post: 01-18-2015, 02:11 PM
  5. Among Europeans, who have "east asian"?
    By Tolan in forum Autosomal (auDNA)
    Replies: 2
    Last Post: 11-25-2013, 09:36 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •