[BioC] A question about lowess normalization

[michael watson (IAH-C)]

Quite simply, your data contains NA values.
These can occur if your background subtracted values are <= 0 in one of
the channels, leading to a nonsensical log(ratio) value.
If you have a lot of NA values, this can be a problem, especially for
print tip loess, as each print tip group has less values to work with,
if lots of them are NA values then the loess fit may not be
representative of your data.  I have seen a log(ratio) switch from -5 to
+5 due to this problem.

To reduce the number of NA values you have, try one of the following: do
not subtract background; use kooperberg background correction; subtract
background and then set all values below a threshold to that threshold.

Mick

-----Original Message-----
From: bioconductor-bounces at stat.math.ethz.ch
[mailto:bioconductor-bounces at stat.math.ethz.ch] On Behalf Of Tony Kowal
Sent: 06 December 2004 02:56
To: bioconductor at stat.math.ethz.ch
Subject: [BioC] A question about lowess normalization


Hi All,

I have just started using R / Bioconductor within the last couple of 
weeks to analyze our microarray data.  While attempting to do the Print 
Tip Lowess normalization on a particular scan, I received the following 
error, and was unable to proceed with the analysis.  I did a google 
search for the error message, but I am still unclear about what is 
wrong with my data.  I have provided the warning I recieved, as well as 
the list of warning that I requested.  If anyone can provide some 
suggestions, I will be extremely grateful.

There are a fairly large number of NA's in my data.  I am using R 2.0 
with the most recent version of Bioconductor.  Analysis is being done 
on a Gateway PC.

Thanks in advance,
Tony

> Error in simpleLoess(y, x, w, span, degree, parametric, drop.square,
> normalize,  :
>         NA/NaN/Inf in foreign function call (arg 1)
> In addition: There were 50 or more warnings (use warnings() to see the

> first 50)
>>  warnings()
> Warning messages:
> 1: zero-width neighborhood. make span bigger
> 2: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 3: zero-width neighborhood. make span bigger
> 4: zero-width neighborhood. make span bigger
> 5: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 6: zero-width neighborhood. make span bigger
> 7: zero-width neighborhood. make span bigger
> 8: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 9: zero-width neighborhood. make span bigger
> 10: zero-width neighborhood. make span bigger
> 11: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 12: zero-width neighborhood. make span bigger
> 13: zero-width neighborhood. make span bigger
> 14: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 15: zero-width neighborhood. make span bigger
> 16: zero-width neighborhood. make span bigger
> 17: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 18: zero-width neighborhood. make span bigger
> 19: zero-width neighborhood. make span bigger
> 20: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 21: zero-width neighborhood. make span bigger
> 22: zero-width neighborhood. make span bigger
> 23: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 24: zero-width neighborhood. make span bigger
> 25: zero-width neighborhood. make span bigger
> 26: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 27: zero-width neighborhood. make span bigger
> 28: zero-width neighborhood. make span bigger
> 29: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 30: zero-width neighborhood. make span bigger
> 31: zero-width neighborhood. make span bigger
> 32: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 33: zero-width neighborhood. make span bigger
> 34: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 35: zero-width neighborhood. make span bigger
> 36: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 37: zero-width neighborhood. make span bigger
> 38: zero-width neighborhood. make span bigger
> 39: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 40: zero-width neighborhood. make span bigger
> 41: zero-width neighborhood. make span bigger
> 42: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 43: zero-width neighborhood. make span bigger
> 44: zero-width neighborhood. make span bigger
> 45: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 46: zero-width neighborhood. make span bigger
> 47: zero-width neighborhood. make span bigger
> 48: ihat=1 (diag L) in l2fit only makes sense if z=x (eval=data).
> 49: zero-width neighborhood. make span bigger
> 50: zero-width neighborhood. make span bigger



Tony Kowal
Graduate Student
	Department of Cell and Molecular Biology
Northwestern University
Chisholm Lab
303 E. Chicago Ave.
Rm 11-100
Chicago, IL 60611

e-mail: askowal at northwestern.edu

_______________________________________________
Bioconductor mailing list
Bioconductor at stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/bioconductor