R: Find most likely separation between positive and negative...
density1d
R Documentation
Find most likely separation between positive and negative
populations in 1D
Description
The function tries to find a reasonable split point between the two
hypothetical cell populations "positive" and "negative". This function
is considered internal, please use the API provided by
rangeGate.
A character scalar giving the flow parameter for which
to compute the separation.
alpha
A tuning parameter that controls the location of the
split point between the two populations. This has to be a numeric in
the range [0,1], where values closer to 0 will shift the
split point closer to the negative population and values closer to 1
will shift towards the positive population. Additionally, the value
of alpha can be "min", in which case the split point
will be selected as the area of lowest local density between the two
populations.
sd
For the case where there is only a single population, the
algorithm falls back to esitmating the mode of this population and a
robust measure of the variance of it distribution. The sd
tuning parameter controls how far away from the mode the split point
is set.
plot
Create a plot of the results of the computation.
borderQuant
Usualy the instrument is set up in a way that the
positive population is somewhere on the high end of the measurement
range and the negative population is on the low end. This parameter
allows to disregard populations with mean values in the extreme
quantiles of the data range. It's value should be in the range
[0,1].
absolute
Logical controling whether to classify a population
(positive or negative) relative to the theoretical measurment range
of the instrument or the actual range of the data. This can be set
to TRUE if the alignment of the measurment range is not
optimal and the bulk of the data is on one end of the theoretical
range.
inBetween
Force the algorithm to put the separator in between
two peaks. If there are more than two peaks, this argument is
ignored.
refLine
Either NULL or a numeric of lenth 1. If
NULL, this parameter is ignored. When it is set to a numeric, the
minor sub-population (if any) below this reference line
will be igored while determining the separator between positive and
negative.
rare
Either TRUE or FALSE, assumes that there is one major peak, and that the rare positive population is to the right of it. Uses a robust estimate of mean and variance to gate the positive cells.
bwFac
The bandwidth for smoothing the density estimate. User-tunable
sig
a value of c(NULL,"L","R"),when sig is not NULL,use the half (left or right) of signal to estimate the std and mean.
peakNr
when peakNr is not NULL,drop the less significant peaks by their heights
...
Further arguments.
Details
The algorithm first tries to identify high density regions in the
data. If the input is a flowSet, density regions will be
computed on the collapsed data, hence it should have been normalized
before (see warpSet for one possible normalization
technique). The high density regions are then clasified as positive
and negative populations, based on their mean value in the theoretical
(or absolute if argument absolute=TRUE) measurement range. In
case there are only two high-density regions the lower one is usually
clasified as the negative populations, however the heuristics in the
algorithm will force the classification towards a positive population
if the mean value is already very high. The absolute and
borderQuant arguments can be used to control this
behaviour. The split point between populations will be drawn at the
value of mimimum local density between the two populations, or, if the
alpha argument is used, somewhere between the two populations
where the value of alpha forces the point to be closer to the negative
(0 - 0.5) or closer to the positive population (0.5 -
1).
If there is only a single high-density region, the algorithm will fall
back to estimating the mode of the distribution
(hubers) and a robust measure of it's variance
and, in combination with the sd argument, set the split point
somewhere in the right or left tail, depending on the classification
of the region.
For more than two populations, the algorithm will still classify each
population into positive and negative and compute the split point
between those clusteres, similar to the two population case.
Value
A numeric indicating the split point between positive and negative populations.
Author(s)
Florian Hahne
See Also
warpSet, rangeGate
Examples
data(GvHD)
dat <- GvHD[pData(GvHD)$Patient==10]
dat <- transform(dat, "FL4-H"=asinh(`FL4-H`), "FL3-H"=asinh(`FL3-H`))
d <- flowStats:::density1d(dat, "FL4-H", plot=TRUE)
if(require(flowViz))
densityplot(~`FL4-H`, dat, refline=d)
## tweaking the location
flowStats:::density1d(dat, "FL4-H", plot=TRUE, alpha=0.8)
## only a single population
flowStats:::density1d(dat, "FL3-H", plot=TRUE)
flowStats:::density1d(dat, "FL3-H", plot=TRUE, sd=2)
Results
R version 3.3.1 (2016-06-21) -- "Bug in Your Hair"
Copyright (C) 2016 The R Foundation for Statistical Computing
Platform: x86_64-pc-linux-gnu (64-bit)
R is free software and comes with ABSOLUTELY NO WARRANTY.
You are welcome to redistribute it under certain conditions.
Type 'license()' or 'licence()' for distribution details.
R is a collaborative project with many contributors.
Type 'contributors()' for more information and
'citation()' on how to cite R or R packages in publications.
Type 'demo()' for some demos, 'help()' for on-line help, or
'help.start()' for an HTML browser interface to help.
Type 'q()' to quit R.
> library(flowStats)
Loading required package: flowCore
Loading required package: fda
Loading required package: splines
Loading required package: Matrix
Attaching package: 'Matrix'
The following object is masked from 'package:flowCore':
%&%
Attaching package: 'fda'
The following object is masked from 'package:graphics':
matplot
Loading required package: mvoutlier
Loading required package: sgeostat
sROC 0.1-2 loaded
Loading required package: cluster
Loading required package: flowWorkspace
Loading required package: flowViz
Loading required package: lattice
Loading required package: ncdfFlow
Loading required package: RcppArmadillo
Loading required package: BH
Loading required package: gridExtra
> png(filename="/home/ddbj/snapshot/RGM3/R_BC/result/flowStats/density1d.Rd_%03d_medium.png", width=480, height=480)
> ### Name: density1d
> ### Title: Find most likely separation between positive and negative
> ### populations in 1D
> ### Aliases: density1d
>
> ### ** Examples
>
>
> data(GvHD)
> dat <- GvHD[pData(GvHD)$Patient==10]
> dat <- transform(dat, "FL4-H"=asinh(`FL4-H`), "FL3-H"=asinh(`FL3-H`))
> d <- flowStats:::density1d(dat, "FL4-H", plot=TRUE)
> if(require(flowViz))
+ densityplot(~`FL4-H`, dat, refline=d)
>
> ## tweaking the location
> flowStats:::density1d(dat, "FL4-H", plot=TRUE, alpha=0.8)
[1] 3.705714
>
> ## only a single population
> flowStats:::density1d(dat, "FL3-H", plot=TRUE)
[1] 4.53325
> flowStats:::density1d(dat, "FL3-H", plot=TRUE, sd=2)
[1] 4.53325
>
>
>
>
>
>
> dev.off()
null device
1
>