Rama,
Feel free to send me your plots. If your distribution of the means
(first row of plots) are approximately symmetric and bell shaped and
the second row is skewed but not in an overly extreme way---and if
there is no bimodality in these plots--then you will likely not see
any difference between the parametric and non-parametric adjustments.
Hope this helps,
Evan
On Jun 9, 2014, at 1:05 PM, Rama Akondy
<rama.akondy1@gmail.com<mailto:rama.akondy1@gmail.com>> wrote:
Hi, I dont much experience analyzing microarray data and this is a
request for some information regarding ComBat that I used
inGenePattern, basically to see if I have gone about it correctly.
In the new documentation (which is very helpful) : there is a sentence
that describes whether to use the parametric or non-parametric method
"In the plots, if the black (kernal density estimate of batch effects)
and red (parametric estimate of batch effects) lines do not overlap,
then the non-parametric method should be used."
I see 4 plots in my Combat output (using the default parametric
method) , two of them seem density plots and two Q-Q plots. The lines
over in one density plot and not in the other. Please let me know how
to interpret this.
Thanks
Rama
Rama S.Akondy, Ph.D
Emory University
[[alternative HTML version deleted]]
Hi, I don't much experience analyzing microarray data and this is a
request
for some information regarding ComBat that I used inGenePattern,
basically
to see if I have gone about it correctly.
In the new documentation (which is very helpful) : there is a sentence
that
describes whether to use the parametric or non-parametric method "In
the
plots, if the black (kernal density estimate of batch effects) and red
(parametric estimate of batch effects) lines do not overlap, then the
non-parametric method should be used."
I see 4 plots in my Combat output (using the default parametric
method) ,
two of them seem density plots and two Q-Q plots. The lines over in
one
density plot and not in the other. Please let me know how to interpret
this.
Thanks
Rama
Rama S.Akondy, Ph.D
Emory University
[[alternative HTML version deleted]]
Evan,
Thank you so much for the prompt response. The second row looks skewed
and
bimodal (plots attached), would you agree? I tried the non-parametric
method
yesterday but basically, the job never finished - I guess that's a
gene
pattern issue.
Many thanks again
Rama
From: Johnson, William Evan [mailto:wej@bu.edu]
Sent: Monday, June 09, 2014 1:33 PM
To: Rama Akondy
Cc: bioconductor at r-project.org
Subject: Re: ComBat
Rama,
Feel free to send me your plots. If your distribution of the means
(first
row of plots) are approximately symmetric and bell shaped and the
second row
is skewed but not in an overly extreme way---and if there is no
bimodality
in these plots--then you will likely not see any difference between
the
parametric and non-parametric adjustments.
Hope this helps,
Evan
On Jun 9, 2014, at 1:05 PM, Rama Akondy <rama.akondy1 at="" gmail.com="">
wrote:
Hi, I don't much experience analyzing microarray data and this is a
request
for some information regarding ComBat that I used inGenePattern,
basically
to see if I have gone about it correctly.
In the new documentation (which is very helpful) : there is a sentence
that
describes whether to use the parametric or non-parametric method "In
the
plots, if the black (kernal density estimate of batch effects) and red
(parametric estimate of batch effects) lines do not overlap, then the
non-parametric method should be used."
I see 4 plots in my Combat output (using the default parametric
method) ,
two of them seem density plots and two Q-Q plots. The lines over in
one
density plot and not in the other. Please let me know how to interpret
this.
Thanks
Rama
Rama S.Akondy, Ph.D
Emory University
