PeakSegFPOP_vec {PeakSegDisk} | R Documentation |
Convert integer data vector to run-length encoding,
then run PeakSegFPOP_df
.
PeakSegFPOP_vec(count.vec, pen.num)
count.vec |
integer vector, noisy non-negatve count data to segment. |
pen.num |
Non-negative numeric scalar. |
List of solver results, same as PeakSegFPOP_dir
.
Toby Dylan Hocking
## Simulate a sequence of Poisson data. sim.seg <- function(seg.mean, size.mean=15){ seg.size <- rpois(1, size.mean) rpois(seg.size, seg.mean) } set.seed(1) seg.mean.vec <- c(1.5, 3.5, 0.5, 4.5, 2.5) z.list <- lapply(seg.mean.vec, sim.seg) z.rep.vec <- unlist(z.list) ## Plot the simulated data. library(ggplot2) count.df <- data.frame( position=seq_along(z.rep.vec), count=z.rep.vec) gg.count <- ggplot()+ geom_point(aes( position, count), shape=1, data=count.df) gg.count ## Plot the true changepoints. n.segs <- length(seg.mean.vec) seg.size.vec <- sapply(z.list, length) seg.end.vec <- cumsum(seg.size.vec) change.vec <- seg.end.vec[-n.segs]+0.5 change.df <- data.frame( changepoint=change.vec) gg.change <- gg.count+ geom_vline(aes( xintercept=changepoint), data=change.df) gg.change ## Fit a peak model and plot it. fit <- PeakSegDisk::PeakSegFPOP_vec(z.rep.vec, 10.5) gg.change+ geom_segment(aes( chromStart+0.5, mean, xend=chromEnd+0.5, yend=mean), color="green", data=fit$segments) ## A pathological data set. z.slow.vec <- 1:length(z.rep.vec) fit.slow <- PeakSegDisk::PeakSegFPOP_vec(z.slow.vec, 10.5) rbind(fit.slow$loss, fit$loss)