Last updated: 2018-05-18

workflowr checks: (Click a bullet for more information)

  • ✔ R Markdown file: up-to-date

    Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.

  • ✔ Environment: empty

    Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.

  • ✔ Seed: set.seed(20180501)

    The command set.seed(20180501) was run prior to running the code in the R Markdown file. Setting a seed ensures that any results that rely on randomness, e.g. subsampling or permutations, are reproducible.

  • ✔ Session information: recorded

    Great job! Recording the operating system, R version, and package versions is critical for reproducibility.

  • ✔ Repository version: 585d9b1

    Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility. The version displayed above was the version of the Git repository at the time these results were generated.

    Note that you need to be careful to ensure that all relevant files for the analysis have been committed to Git prior to generating the results (you can use wflow_publish or wflow_git_commit). workflowr only checks the R Markdown file, but you know if there are other scripts or data files that it depends on. Below is the status of the Git repository when the results were generated: 
    
Ignored files:
    Ignored:    .Rhistory
    Ignored:    .Rproj.user/
    Ignored:    log/

Untracked files:
    Untracked:  analysis/binom.Rmd
    Untracked:  analysis/overdis.Rmd
    Untracked:  analysis/smashtutorial.Rmd
    Untracked:  data/treas_bill.csv
    Untracked:  docs/figure/smashtutorial.Rmd/

Unstaged changes:
    Modified:   analysis/ashpmean.Rmd
    Modified:   analysis/nugget.Rmd
    Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.
Expand here to see past versions:
    File Version Author Date Message
    Rmd 585d9b1 Dongyue 2018-05-18 binomial smooth vs poibinom

Methods

Let \(X\sim Binomial(n,p)\) then \(E(X)=np, Var(X)=np(1-p)\). Poisson distribution is an approximation of binomial distribution when \(n\) is large and \(p\) is small. A rule of thumb is that \(n\geq 20, p\leq 0.05\).

Derivation: Let \(\lambda=np\)

\[\frac{n!}{x!(n-x)!}p^x(1-p)^{n-x}=\frac{n(n-1)...(n-k+1)}{x!}(\lambda/n)^x(1-\lambda/n)^{n-x}\approx \frac{\lambda^x}{x!}(1-\lambda/n)^{n-x}\] as \(n\to \infty\). Since \(lim_{n\to \infty}(1-\lambda/n)^{n}=e^{-\lambda}\) and \(lim_{n\to \infty}(1-\lambda/n)^{-x}=1\), we have \[\frac{n!}{x!(n-x)!}p^x(1-p)^{n-x}\approx \frac{\lambda^x e^{-\lambda}}{x!}.\]

If we have binomial observation \(X_t\) with \(n_t\) and treat it as Poisson observation, we can do the following expansion: \[Y_t=\log(X_t)=\log(n_tp_t)+\frac{X_t-n_tp_t}{n_tp_t}=\log(n_t)+\log(p_t)+\frac{X_t-n_tp_t}{n_tp_t}.\] This leads to \[Y_t-\log(n_t)=\log(p_t)+\frac{X_t-n_tp_t}{n_tp_t}.\]

Experiments

We compare the performance of smashgen - binomial and smashgen - poi_binom, as well as Translation Invariant (TI) thresholding (Coifman and Donoho, 1995), which is one of the best methods in a large-scale simulation study in Antoniadis et al. (2001), and Ebayesthresh (Johnstone and Silverman, 2005b).

For all experiments, T is set to be 256, nugget effect \(\sigma=1\). The mean squared errors are reported and the plots are served as visual aids.

library(smashrgen)
library(ggplot2)
library(EbayesThresh)

simu_study=function(p,sigma=1,ntri,nsimu=100,seed=12345,
                    niter=1,family='DaubExPhase',ashp=TRUE,verbose=FALSE,robust=FALSE,
                    tol=1e-2){
  set.seed(seed)
  smash.binom.err=c()
  smash.poibinom.err=c()
  ti.thresh.err=c()
  eb.thresh.err=c()
  n=length(p)
  target=exp(p)/(1+exp(p))
  for(k in 1:nsimu){
    ng=rnorm(n,0,sigma)
    m=exp(p+ng)
    q=m/(1+m)
    x=rbinom(n,ntri,q)
    #fit data
    smash.binom.out=smash_gen(x,dist_family = 'binomial',y_var_est='smashu',ntri=ntri)
    smash.poibinom.out=smash_gen(x,dist_family = 'poi_binom',y_var_est='smashu',ntri=ntri)
    ti.thresh.out=ti.thresh(x/ntri,method='rmad')
    eb.thresh.out=waveti.ebayes(x/ntri)
    #errors
    smash.binom.err[k]=mse(target,smash.binom.out)
    smash.poibinom.err[k]=mse(target,smash.poibinom.out)
    ti.thresh.err[k]=mse(target,ti.thresh.out)
    eb.thresh.err[k]=mse(target,eb.thresh.out)
  }
  return(list(est=list(smash.binom.out=smash.binom.out,smash.poibinom.out=smash.poibinom.out, ti.thresh.out=ti.thresh.out,eb.thresh.out=eb.thresh.out,x=x),
              err=data.frame(smash.binom=smash.binom.err,smash.poibinom=smash.poibinom.err, ti.thresh=ti.thresh.err,eb.thresh=eb.thresh.err)))
}

waveti.ebayes = function(x, filter.number = 10, family = "DaubLeAsymm", min.level = 3) {
    n = length(x)
    J = log2(n)
    x.w = wd(x, filter.number, family, type = "station")
    for (j in min.level:(J - 1)) {
        x.pm = ebayesthresh(accessD(x.w, j))
        x.w = putD(x.w, j, x.pm)
    }
    mu.est = AvBasis(convert(x.w))
    return(mu.est)
}

Constant trend

\(ntri\) large, \(p\) large

The number of trials are generated from a Poisson distribution with \(\lambda=50\). \(p\) is around 0.8.

n=256
p=rep(1.5,n)
set.seed(111)
ntri=rpois(n,50)
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

\(ntri\) small, \(p\) large

The number of trials are generated from a Poisson distribution with \(\lambda=8\). \(p\) is around 0.8.

n=256
p=rep(1.5,n)
set.seed(111)
ntri=rpois(n,5)+1
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

\(ntri\) large, \(p\) small

The number of trials are generated from a Poisson distribution with \(\lambda=50\). \(p\) is around 0.05.

n=256
p=rep(-3,n)
set.seed(111)
ntri=rpois(n,50)
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

ggplot(df2gg(result$err[,1:2]),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

As expected, when \(n\) is large and \(p\) is small, Poisson distribution is a good approximation to binomial distribution.

\(ntri\) small, \(p\) small

The number of trials are generated from a Poisson distribution with \(\lambda=5\). \(p\) is around 0.05.

n=256
p=rep(-3,n)
set.seed(111)
ntri=rpois(n,5)+1
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

Steps

Large \(n\)

p=c(rep(-2,64), rep(0, 64), rep(2, 64), rep(-2, 64))
set.seed(111)
ntri=rpois(256,50)
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

Small \(n\)

p=c(rep(-2,64), rep(0, 64), rep(2, 64), rep(-2, 64))
set.seed(111)
ntri=rpois(256,5)+1
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

Bumps

Large \(n\).

m=seq(0,1,length.out = 256)
h = c(4, 5, 3, 4, 5, 4.2, 2.1, 4.3, 3.1, 5.1, 4.2)
w = c(0.005, 0.005, 0.006, 0.01, 0.01, 0.03, 0.01, 0.01, 0.005,0.008,0.005)
t=c(.1,.13,.15,.23,.25,.4,.44,.65,.76,.78,.81)
f = c()
for(i in 1:length(m)){
  f[i]=sum(h*(1+((m[i]-t)/w)^4)^(-1))
}
p=f-3

set.seed(111)
ntri=rpois(256,50)
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

Small \(n\)

set.seed(111)
ntri=rpois(256,5)+1
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

Spike mean

Large \(n\)

spike.f = function(x) (0.75 * exp(-500 * (x - 0.23)^2) + 1.5 * exp(-2000 * (x - 0.33)^2) + 3 * exp(-8000 * (x - 0.47)^2) + 2.25 * exp(-16000 * 
    (x - 0.69)^2) + 0.5 * exp(-32000 * (x - 0.83)^2))
n = 256
t = 1:n/n
p = spike.f(t)*2-2

set.seed(111)
ntri=rpois(256,50)
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

Small \(n\)

set.seed(111)
ntri=rpois(256,5)+1
result=simu_study(p,ntri=ntri)

ggplot(df2gg(result$err),aes(x=method,y=MSE))+geom_boxplot(aes(fill=method))+labs(x='')

par(mfrow=c(2,2))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.binom.out,col=4,lwd=2)
legend("bottomright", # places a legend at the appropriate place
       c("truth","smash-binom"), # puts text in the legend
       lty=c(2,1), # gives the legend appropriate symbols (lines)
       lwd=c(1,2),
       cex = 1,
       col=c("black","blue"))
plot(result$est$x/ntri,col='gray80',ylab='',main='smash-poi_binom')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$smash.poibinom.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='TI thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$ti.thresh.out,col=4,lwd=2)
plot(result$est$x/ntri,col='gray80',ylab='',main='EBayes thresh')
lines(exp(p)/(1+exp(p)),lty=2)
lines(result$est$eb.thresh.out,col=4,lwd=2)

Session information

sessionInfo()
R version 3.4.0 (2017-04-21)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 16299)

Matrix products: default

locale:
[1] LC_COLLATE=English_United States.1252 
[2] LC_CTYPE=English_United States.1252   
[3] LC_MONETARY=English_United States.1252
[4] LC_NUMERIC=C                          
[5] LC_TIME=English_United States.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] EbayesThresh_1.4-12 ggplot2_2.2.1       smashrgen_0.1.0    
[4] wavethresh_4.6.8    MASS_7.3-47         caTools_1.17.1     
[7] ashr_2.2-7          smashr_1.1-5       

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.16        plyr_1.8.4          compiler_3.4.0     
 [4] git2r_0.21.0        workflowr_1.0.1     R.methodsS3_1.7.1  
 [7] R.utils_2.6.0       bitops_1.0-6        iterators_1.0.8    
[10] tools_3.4.0         digest_0.6.13       tibble_1.3.3       
[13] evaluate_0.10       gtable_0.2.0        lattice_0.20-35    
[16] rlang_0.1.2         Matrix_1.2-9        foreach_1.4.3      
[19] yaml_2.1.19         parallel_3.4.0      stringr_1.3.0      
[22] knitr_1.20          REBayes_1.3         rprojroot_1.3-2    
[25] grid_3.4.0          data.table_1.10.4-3 rmarkdown_1.8      
[28] magrittr_1.5        whisker_0.3-2       backports_1.0.5    
[31] scales_0.4.1        codetools_0.2-15    htmltools_0.3.5    
[34] assertthat_0.2.0    colorspace_1.3-2    labeling_0.3       
[37] stringi_1.1.6       Rmosek_8.0.69       lazyeval_0.2.1     
[40] munsell_0.4.3       doParallel_1.0.11   pscl_1.4.9         
[43] truncnorm_1.0-7     SQUAREM_2017.10-1   R.oo_1.21.0

This reproducible R Markdown analysis was created with workflowr 1.0.1