Numerical error: a or b too big, or MAXIT too small in betac

[Cheng Li]

Posted: 26 Dec 2007 02:48 pm Post subject: Numerical error: a or b
too big, or MAXIT too small in betac

--------------------------------------------------------------------------------

When I was trying to run a ANOVA test in dchip, got the following
error:

{ANOVA or correlation filtering
Obtain factor information...
Use sample information column 'Profile '
90 of 90 samples have information on 'Profile '

Filtering by p-value 0.050000
Use array list file 'C:\bioinfo\dchip\dChip_array_list.txt', use
standardize separators: Yes
Gene 0
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf


What does it mean?



Posted: 26 Dec 2007 05:35 pm Post subject:

--------------------------------------------------------------------------------

how do your sample info file and array list file look like?

you can also try unchecking "tools/array list file/use standardize
separators".

cheng



Posted: 26 Dec 2007 06:01 pm Post subject:

--------------------------------------------------------------------------------

uncheck the "separator" does not help.

Sample Info file is like:
HG-U133A Stanley ID Filename Collection Type Age *** Race Profile DSM
IV Axis I Primary Diagnosis PMI Brain pH Left Brain Suicide Status
Psychotic Feature Rate of Death Exacerbation Smoking at Time of Death
Lifetime Alcohol Lifetime Drugs Lifetime Anti-psychotics
U133A0303 Stanley FG 7 A-7 U133A0303 Array 30 M White Schizophrenia
295.3 SCHIZ PARANOID 33 6.2 Fixed No Yes Sudden Probably Not Yes Heavy
in present Heavy in past 35000
U133A0304 Stanley FG 3 A-3 U133A0304 Array 45 M White Schizophrenia
295.9 SCHIZ UNDIFF 26 6.42 Frozen No Yes Possible anoxia Probably Not
No Little or none Little or none 180000


Array list is like:
A-7
A-3
A-15
A-1
A-5
A-9
A-82
A-27
A-24
A-89
A-54

I did not use Array list, still got the same error.



Posted: 28 Dec 2007 12:09 am Post subject: ANOVA issue

--------------------------------------------------------------------------------

could you try to use this 12/27 dchip:
http://www.hsph.harvard.edu/~cli/dchip.exe

and follow the steps below to create a data package so I can check the
issue?

1. after "Open group", use "Tools/export expression value" to export a
data file. you can select a subset of samples as long as the ANOVA
issue remains.

2. use "analysis/get external data" to read in this exported data
file, specifying the sample information file, do not specify gene info
file.

3. go to "tools/automate", specify an "automation file name", add the
menu step "get external data", click "Save & Exit".

4. perform "Analysis/ANOVA" so that it reports the error.

5. go to "tools/automate", add the menu step "analysis/ANOVA", select
both menu steps, click "Save & Exit".

6. go to "tools/automate", click "Package data". a new data directory
will be created. go to this directory and click "dchip.exe" to run the
above two menu steps so that it reports error.

7. zip all the files in this directory into a zip file and send it to
me by email or via www.transferbigfiles.com.

cheng



Posted: 28 Dec 2007 06:45 pm Post subject:

--------------------------------------------------------------------------------

several new problems:
1. why the sample information file has to be located in the same
directory of dchip?
When I placed it in a different directory, dchip just can't open it.

2. I tried to follow your instruction, exporting data and re-importing
it.
The header and gene annotation data in the exported data is quite
annoying.
Always gave error message like "Absolute calls may be in an expression
data colun; please .....", till I have to manually delete those gene
annotation columns which I did not intent to export -- I don't
understand where is the control for that in the exporting step.

Anyway, I managed to go through your steps without knowing what those
"automate" steps mean.
I will upload the zipped data to you, hope it will work.



Posted: 29 Dec 2007 12:52 am Post subject:

--------------------------------------------------------------------------------

Thanks for the data files. This is caused by probe sets with many
missing or truncated expression values (to 1). You can use this
12/28/07 dchip to skip these probe sets by setting their p-values to
1:
Http://biosun1.harvard.edu/~cli/dchip.exe

1) yes the new dchip requires sample info file to be in the working
directory.

2) if you don't specify gene info file at "open group", gene
annotation will not be exported at "tools/export expression". or you
can specify "skip columns" at "get external data".

Cheng



Posted: 29 Dec 2007 12:58 am Post subject:

--------------------------------------------------------------------------------

The "Brain pH" is not a categorical variable, you better set it as
"Brain pH(nuermic)" and use correlation fitlering.

you can only use up to two factors in the ANOVA menu. if still in
error you can post dchip output.

cheng

Posted: 28 Dec 2007 07:12 pm Post subject: ANOVA test only one factor
each time?

when trying to run ANOVA test, if I select more than one "factors," it
will give error like no sample information column named "Profile ,
pH."
Actually, these are two factors when I select them together for
ANOVA.



Posted: 12 Jun 2008 03:19 pm Post subject:

--------------------------------------------------------------------------------

Xinmin,

This due to truncated expression value 1 across all samples in a probe
set, leading to error in computing the standard error of the group
mean.

Please try this updated version: http://biosun1.harvard.edu/~cli/dchip.exe

Thanks for the message,

Cheng Li

________________________________________
From: Li, Xinmin
Sent: Thursday, June 12, 2008 1:41 PM
To: Cheng Li
Subject: RE: dchip normalization for external data

Hi Cheng,

Sorry to bother you. I am planning a dChip workshop at UCLA now.
Realize the software has change a bit. Here is another question:

When I perform comparison analysis, I can do it without any warning if
I do not specify p value. however, if I specify p value, I got an
error message as below:

Writing comparison result in 'C:\UCLA folder 11-6-07\workshop &
training\workshop data analysis practice folder\dChip
workshop_combined compare_result.xls' (file pathname is in
clipboard)...
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
Numerical error: a or b too big, or MAXIT too small in betacf
48 genes satisfied the comparison filtering criteria

Any thought on this error message?

Thanks

xinmin