FW: Lab fun? Bad project video...
Deitiker, Philip R
(by pdeitik from bcm.edu)
Thu Jan 27 17:38:14 EST 2011
From: Deitiker, Philip R
Sent: Thursday, January 27, 2011 4:11 PM
To: 'mailto:Methods from net.bio.net'
Subject: RE: Lab fun? Bad project video...
I suppose that the videos from the halloween party would not help our cause, but anyway, this is not 'Baylor Med' as I know it (at least in the lab, outside of the lab, well....). Though we have some rather 'precocious' graduate students as graduate students tend to be sometimes (too many hours in the lab, not enough sleep, too many cuppachinos, excessive amounts of pizza-hut pizza, not enough fiber, that kind of thing).
Think about the following question.
If a distribution of data points is skewed (2.5) and kurtotic (2.5) as a consequence of heteroscedasticity, is it better to present the mean and the standard deviation or the geometric mean +1 Sd/ - 1Sd? Or would it be better to present the standard error of the mean or the 96% confidence interval of geometric SEM?
If one has a cyclical trend how many degrees of freedom should be assessed to the cyclical trend (knowing that any cyclical trend will have an initial theta (a la sine), a period, and an amplitude) and subtracted from the residual degrees of freedom. On top of that if one does regression analysis of the cyclical residuals, how many should one treat the regression degrees of freedom when calculating the F statistic and computing the probability distribution that could give rise to the F statistic? If there is an increase in the F statistic but this does not result in a lower probability p(F) should the second regression consider to be significant or insignificant? What if the second regression is discontinuous, it increases for a period but then saturates and exhibits no further increase? Should the variance probability of the regression only be considered for the period in which it is increasing?
Secondary if a GWAS association has a probability of 1E-8 but the number of SNP sampled is 5E-6 and it is well known that many genetic markers in the GWAS are equivilant (as a consequence of very strong linkage disequilibrium). Is it a legitimate correction to divide the 1E-8/5E-6 to arrive at a corrected probability of 0.002, or is better to correct by some assessment of haplotype diversity and haplotype boundaries within physically grouped SNPs. Suppose you survey a sight in the HLA locus and find a very high association between a marker at a highly heterogenous locus (e.g. HLA-DRB1), but the GWAS shows insignificantly or marginally low probabilities at this locus, although GWAS cannot possibly survey all the heterogeneity at that locus (number of alleles exceeds the Hap-Map sample size for the region), how does one compare the results of haplotype associations of the GWAS SNP allele (meaning one SNP) with the local results for a given allele (meaning entire coding frame) association?
OK so welcome to my world here at BCM. Not to mention I had a mock inspection this morning, prior to the real AAALAC nightmare that will take place over the next 2 months (given the last visit, no-understatement), the fact they laid off 1000+ personal at the end of last year, and a broad variety of other compliance (CDC, Patriot act, Animal welfare act, radioactivity, chemical inventory, high-rise fire code compliance, blood born pathogen, etc) and 'right-sizing' hair-pullers (and I am not talking about capital pilitation)?
Exciting, right, make a video about it, yeah?
Life after the Ph.D. at BCM from an ex-graduate student at BCM.
From: methods-bounces from oat.bio.indiana.edu [mailto:methods-bounces from oat.bio.indiana.edu] On Behalf Of DK
Sent: Wednesday, January 26, 2011 11:38 PM
To: methods from magpie.bio.indiana.edu
Subject: Re: Lab fun? Bad project video...
In article <a75d37f8-4c27-4c96-859a-302d995eea32 from s5g2000yqm.googlegroups.com>, WS <novalidaddress from nurfuerspam.de> wrote:
I am only surprised it's not yet at 1,000,000 views.
Don't know anyone who hasn't seen it :-)
My favorite part is "what, what, what, what's in this box"
(at 2:38). The Craigslist job listing in North Dakota
at 4:36 is a great touch too.
Methods mailing list
Methods from net.bio.net
More information about the Methods