[RASMB] "wavy" interference optics residuals (final bit of detail)

[Arthur Rowe]

There is nothing like stumbling into one's own heffalump trap, is there? As
has been well pointed out by Peter Schuck and others (including myself, I
have to own up), the error surface for fitting multiple parameters to SE
data is very lumpy indeed. So -believing the output from a simple single
Levenberg-Marquadt fit is a hazardous thing to do. So - YES - that BM =
0.001498 value for the fit to wave noise added data IS an outlier, which by
pure coincidence is close to the true answer.

Doing it properly, full error analysis, 200 iterations 68.3% confidence
intervals, returns BM = 1.00142 ±0.0004. Much more realistic. But still not
at all bad an answer for what we are trying to achieve.

Regards (again)

Arthur



Hi Debra (again)

Checking back my records - which I should have done before mailing - I see
that the frequency of the wave noise I saw was only around 1 cycle (2¼) in 2
mm radial distance.

I have quickly done a computer simulation of the situation, taking a
non-ideal system with sigma(monomer) = 1.5, BM = 0.0015 (fringe units)
loading concentration 3 fringes (~1 mg/ml), with ±0.007 SD of stochastic
noise added to each data point.  Floating BM, c(reference) and baseline (E)
without wave noise, fitting* gave an estimate for BM = 0.001501. Whereas
doing the same fit with 1 cycle of wave noise added (amplitude 0.005)
returned a value BM = 0.001498!

So - given that for a globular protein at ~1 mg/ml you are pushing it to get
a reliable value for the second virial term (BM), I think that for most
purposes wave noise of this level of amplitude can be tolerated, mysterious
though it may be. But what do you reckon IS the amplitude of your wave?
Actually I would predict that for 7 cycles within the domain of the data set
you are better off than the situation which I have simulated, since 1 noise
cycle (sine wave) is skewed and thus in not an un-biased addition

Regards

Arthur

*using our INVEQ software

Hi All, 

We have been experiencing a problem with our interference data in which
after data fitting, the residuals frequently have a systematic ³wavy² look
to them where they rise above and below zero multiple times >7 across the
fitted data.  Subtracting a water blank does not fix the problem, and it has
occurred an several different samples.  In real time, the fringes also
³jump² during centrifugation.  Is this an indication of an equipment
problem, perhaps that there is something unstable in the centrifuge and/or
optical system that is leading to the movement of the fringe pattern
detected by the camera?  If so, is there a likely culprit?  If not, what is
likely to cause this?

Thanks for the help,

Debra M. Eckert, Ph.D.

Research Assistant Professor of Biochemistry

University of Utah 

deckert at biochem.utah.edu

 

 

_______________________________________________
RASMB mailing list
RASMB at rasmb.bbri.org
http://rasmb.bbri.org/mailman/listinfo/rasmb




This message has been checked for viruses but the contents of an attachment
may still contain software viruses, which could damage your computer system:
you are advised to perform your own checks. Email communications with the
University of Nottingham may be monitored as permitted by UK legislation.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.rasmb.org/pipermail/rasmb-rasmb.org/attachments/20090717/c6ea6693/attachment.htm>