Don't run 500K separate models. Use the limma package to fit one model that
can learn the variance parameters jointly. Run it on your laptop. And don't
use %methylation as your Y variable, use logit(percent), i.e. the Beta
value.
-Aaron
On Mon, Aug 8, 2016 at 2:49 PM, Ellis, Alicia M
wrote:
> I
I'm also curious how to use glmnet with survfit -- specifically, for use
with interval regression (which, under the hood, is implemented using
survfit). Can you show how you converted your Surv object formula to a
design matrix for use with glmnet?
Thanks,
-Aaron
On Sun, Dec 8, 2013 at 12:45 AM
On Fri, Sep 20, 2013 at 10:10 AM, Preetam Pal wrote:
> I have 25 variables in the data file (name: score), i.e. X1,X2,.,X25.
>
> I dont want to use score$X1, score$X2 everytime I use these variables.
>
attach(score)
plot(X1, X2) # etc. etc.
-Aaron
[[alternative HTML version delete
for plotting purposes, I typically jitter() the x's and y's to see the
otherwise overlapping data points
-Aaron
On Wed, Jun 26, 2013 at 12:29 PM, Shane Carey wrote:
> Nope, neither work. :-(
>
>
> On Wed, Jun 26, 2013 at 5:16 PM, Clint Bowman wrote:
>
> > John,
> >
> > That still leaves a strin
shameless self-plug: we break out of R to do this, and after many painful
years developing and maintaining idiosyncratic Makefiles, we are now using
Taverna to (visually) glue together UNIX commands (including R scripts) --
the benefits of which (over make and brethren) is that you can actually
*se
Without really knowing this code, I can guess that it may be the
"triangular" prior at work. Bayes Factors are notorious for being sensitive
to the prior. Presumably, the prior somehow prefers to see the rarer allele
as the "BB", and not the "AA" homozygous genotype (this is a common
assumption:
H-W only gives you the expected frequency of AA, AB, and BB genotypes (i.e.
a 1x3 table):
minor <- runif(1, 0.05, 0.25)
major <- 1-minor
AA <- minor^2
AB <- 2*minor*major
BB <- major^2
df <- cbind(AA, AB, BB)
-Aaron
On Tue, Jun 21, 2011 at 9:30 PM, Jim Silverton wrote:
> Hello all,
> I am
You can try something like this, at the command line:
gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.5 -dPDFSETTINGS=/screen
-dNOPAUSE -dQUIET -dBATCH -sOutputFile=output.pdf input.pdf
evidently, the new compactPDF() function in R 2.13 does something very similar.
-Aaron
On Thu, May 19, 2011 at
OR <- exp(coef(GLM.2)[-1])
OR.ci <- exp(confint(GLM.2)[-1,])
-Aaron
On Tue, Mar 15, 2011 at 1:25 PM, lafadnes wrote:
> I am a new R user (am using it through the Rcmdr package) and have
> struggled
> to find out how to report OR and RR directly when running GLM models (not
> only reporting coef
What I think you need is something along the lines of:
matrix(c(sample(3:7), sample(3:7), sample(3:7), sample(3:7), ...), nrow=2)
now, each column are your random pairs.
-Aaron
On Wed, Mar 9, 2011 at 1:01 PM, Hosack, Michael wrote:
> > -Original Message-
> > From: r-help-bounces at r-
FYI, in bioinformatics, we use dynamic programming algorithms in similar
ways to solve similar problems of finding guaranteed-optimal partitions in
streams of data (usually DNA or protein sequence, but sometimes numerical
data from chip-arrays). These "path optimization" algorithms are often
calle
(I think) I'd like to use the hmm.discnp package for a simple discrete,
two-state HMM, but my training data is irregularly shaped (i.e. the
observation chains are of varying length). Additionally, I do not see how
to label the state of the observations given to the hmm() function.
Ultimately, I'd
Thanks to all, "do.call(order, as.data.frame(y))" was the idiom I was
missing!
-Aaron
On Thu, Feb 19, 2009 at 11:52 AM, Gustaf Rydevik
wrote:
> On Thu, Feb 19, 2009 at 5:40 PM, Aaron Mackey wrote:
> > There's got to be a better way to use order() on a matrix than this:
There's got to be a better way to use order() on a matrix than this:
> y
2L-035-3 2L-081-23 2L-143-18 2L-189-1 2R-008-5 2R-068-15 3L-113-4
3L-173-2
3981 1 221 12
2
8571 1 221 22
2
91
nt fashion. If not, I will see if I
> can create/drop the temp table directly from sqlQuery.
> -Avram
>
>
>
> On Thursday, September 11, 2008, at 12:07PM, "Aaron Mackey" <[EMAIL
> PROTECTED]> wrote:
>>Sorry, I see now you want to avoid this, but you d
Sorry, I see now you want to avoid this, but you did ask what was the
"best way to efficiently ...", and the temp. table solution certainly
matches your description. What's wrong with using a temporary table?
-Aaron
On Thu, Sep 11, 2008 at 3:05 PM, Aaron Mackey <[EMAIL PROT
I would load your set of userid's into a temporary table in oracle,
then join that table with the rest of your SQL query to get only the
matching rows out.
-Aaron
On Thu, Sep 11, 2008 at 2:33 PM, Avram Aelony <[EMAIL PROTECTED]> wrote:
>
> Dear R list,
>
> What is the best way to efficiently marr
If you mean you want an EVD with a fat left tail (instead of a fat
right tail), then can;t you just multiply all the values by -1 to
"reverse" the distribution? A new location parameter could then shift
the distribution wherever you want along the number line ...
-Aaron
On Mon, Sep 8, 2008 at 5:
Witness this oddity (to me):
> rainbow_hcl(10)[1]
[1] "#E18E9E"
> d <- attributes(hex2RGB(rainbow_hcl(10)))$coords[1,]
> rgb(d[1], d[2], d[3])
[1] "#C54D5F"
What happened? FYI, this came up as I'm trying to reuse the RGB values I
get from rainbow_hcl in a call to rgb() where I can also set alpha
I have a fairly large model:
> length(Y)
[1] 3051
> dim(covariates)
[1] 3051 211
All of these 211 covariates need to be nested hierarchically within a
grouping "class", of which there are 8. I have an accessory vector, "
cov2class" that specifies the mapping between covariates and the 8 classes
20 matches
Mail list logo
