Power vs. fdr: version 4

Last updated: 2024-01-29

Checks: 7 0

Knit directory: survival-susie/

This reproducible R Markdown analysis was created with workflowr (version 1.6.2). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.

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(20230201)

The command set.seed(20230201) 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.

Cache: none

Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.

File paths: relative

Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.

Repository version: 6130e84

Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.

The results in this page were generated with repository version 6130e84. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.

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:    .Rproj.user/

Untracked files:
    Untracked:  analysis/dat_analysis_cancer_subtype.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.

These are the previous versions of the repository in which changes were made to the R Markdown (analysis/compare_method_large_sample.Rmd) and HTML (docs/compare_method_large_sample.html) files. If you’ve configured a remote Git repository (see ?wflow_git_remote), click on the hyperlinks in the table below to view the files as they were in that past version.

File	Version	Author	Date	Message
Rmd	6130e84	yunqi yang	2024-01-29	wflow_publish("analysis/compare_method_large_sample.Rmd")
html	1c71368	yunqi yang	2024-01-29	Build site.
Rmd	33ea019	yunqi yang	2024-01-29	wflow_publish("analysis/compare_method_large_sample.Rmd")

Description:

Version 4 simulation results, comparing power vs. FDR across 3 methods. I vary the threshold for claiming effect variables based on marginal PIP value.

Difference between v3 and v4 simulation:

Only susie has been changed, not the other two methods.
Used corrected ABF instead of original Wakefeld ABF.
Computed susie credible sets.

Conclusion:

After using corrected ABF, susie performance is improved. Now it’s better than bvsnlp almost all the time.
Power of the credible sets and coverage of credible sets also look fine.


calculate_tpr_vs_fdr <- function(pip, is_effect, ts){
  res <- matrix(NA, nrow = length(ts), ncol = 2)
  colnames(res) = c("tpr", "fdr")
  for (i in 1:length(ts)){
    pred_pos = pip >= ts[i]
    tp = pip >= ts[i] & is_effect == 1
    fp = pip >= ts[i] & is_effect == 0
    tpr = sum(tp)/sum(is_effect)
    fdr = sum(fp)/sum(pred_pos)
    res[i, ] = c(tpr, fdr)
  }
  return(res)
}

# coverage: the proportion of CSs that contain an effect variable
# @param dat_indx: the indx for the data from dsc
calculate_cs_coverage = function(dat_indx){
  contain_status = c()
  for (indx in dat_indx){
     cs = susie$susie.cs[[indx]]$cs
     true_effect = which(susie$simulate.is_effect[[indx]] == 1)
     if (!is.null(cs)){
       for (j in 1:length(cs)){
         res = ifelse(sum(true_effect %in% unlist(cs[j])) ==  1, 1, 0)
         contain_status = c(contain_status, res)
        }
     }
  }
  coverage = sum(contain_status)/length(contain_status)
  return(coverage)
}


# @param dat_indx: the indx for the data from dsc
# @p: number of variables in each simulation replicate. 
get_cs_effect = function(dat_indx, p){
  cs_effect = c()
  for (indx in dat_indx){
    effect = rep(0, p)
    cs_effect_indx = c(unlist(susie$susie.cs[[indx]]$cs))
    effect[cs_effect_indx] = 1
    cs_effect = c(cs_effect, effect)
  }
  return(cs_effect)
}


susie = readRDS("/project2/mstephens/yunqiyang/surv-susie/dsc202401/susie.rds")
susie.rss = readRDS("/project2/mstephens/yunqiyang/surv-susie/dsc202401/susie_rss.rds")
susie.rss.varY = readRDS("/project2/mstephens/yunqiyang/surv-susie/dsc202401/susie_rss_varY.rds")
survsvb = readRDS("/project2/mstephens/yunqiyang/surv-susie/dsc202401/survsvb.rds")
bvsnlp = readRDS("/project2/mstephens/yunqiyang/surv-susie/dsc202401/bvsnlp.rds")
r2b = readRDS("/project2/mstephens/yunqiyang/surv-susie/dsc202401/r2b.rds")

1. Results using real correlation structure from data

par(mfrow = c(2,3), cex.axis = 1.5)
censor_lvl = c(0, 0.2, 0.4, 0.6, 0.8, 0.99)
for (i in 1:length(censor_lvl)){
  indx = which(susie$simulate.cor_type == "real" & susie$simulate.censor_lvl == censor_lvl[i])
  pip.susie = unlist(lapply(indx, function(x) susie$susie.pip[[x]]))
  pip.survsvb = unlist(lapply(indx, function(x) survsvb$svb.pip[[x]]))
  pip.bvsnlp = unlist(lapply(indx, function(x) bvsnlp$bvsnlp.pip[[x]]))
  pip.r2b = unlist(lapply(indx, function(x) r2b$r2b.pip[[x]]))
  is_effect = unlist(lapply(indx, function(x) susie$simulate.is_effect[[x]]))
  
  ts = seq(from = 0, to = 1, by = 0.001)
  res.susie = calculate_tpr_vs_fdr(pip.susie, is_effect, ts)
  res.svb = calculate_tpr_vs_fdr(pip.survsvb, is_effect, ts)
  res.bvsnlp = calculate_tpr_vs_fdr(pip.bvsnlp, is_effect, ts)
  res.r2b = calculate_tpr_vs_fdr(pip.r2b, is_effect, ts)
  
  plot(res.susie[,2], res.susie[,1], type = "l", xlim = c(0,1), ylim = c(0, 1), xlab = "FDR", ylab = "Power",
       main = paste0("Real correlation, effect 0-3", ",censor=", censor_lvl[i]))
  lines(res.svb[,2], res.svb[,1], type = "l", col = 2)
  lines(res.bvsnlp[,2], res.bvsnlp[,1], type = "l", col = 3)
  lines(res.r2b[,2], res.r2b[,1], type = "l", col = 4)
  
  points(res.susie[96,2], res.susie[96, 1])
  points(res.svb[96,2], res.svb[96, 1])
  points(res.bvsnlp[96,2], res.bvsnlp[96, 1])
  points(res.r2b[96,2], res.r2b[96, 1])
  
  legend("topleft", legend = c("susie", "survival.svb", "bvsnlp", "r2b"), col = c(1,2,3, 4), lty = 1)
}

Version	Author	Date
1c71368	yunqi yang	2024-01-29

The dots indicate PIP threshold = 0.95

2. Results using independent X, without data from null model.

par(mfrow = c(2,3),cex.axis = 1.5)
censor_lvl = c(0, 0.2, 0.4, 0.6, 0.8, 0.99)

for (i in 1:length(censor_lvl)){
  indx = which(susie$simulate.cor_type == "independent" & susie$simulate.censor_lvl == censor_lvl[i] & susie$simulate.num_effect != 0)
  
  pip.susie = unlist(lapply(indx, function(x) susie$susie.pip[[x]]))
  pip.survsvb = unlist(lapply(indx, function(x) survsvb$svb.pip[[x]]))
  pip.bvsnlp = unlist(lapply(indx, function(x) bvsnlp$bvsnlp.pip[[x]]))
  pip.r2b = unlist(lapply(indx, function(x) r2b$r2b.pip[[x]]))
  is_effect = unlist(lapply(indx, function(x) susie$simulate.is_effect[[x]]))
  
  ts = seq(from = 0, to = 1, by = 0.001)
  res.susie = calculate_tpr_vs_fdr(pip.susie, is_effect, ts)
  res.svb = calculate_tpr_vs_fdr(pip.survsvb, is_effect, ts)
  res.bvsnlp = calculate_tpr_vs_fdr(pip.bvsnlp, is_effect, ts)
  res.r2b = calculate_tpr_vs_fdr(pip.r2b, is_effect, ts)
  
  
  plot(res.susie[,2], res.susie[,1], type = "l", xlim = c(0, 0.1), ylim = c(0, 1), xlab = "FDR", ylab = "Power",
       main = paste0("X independent, effect 1-3", ",censor=", censor_lvl[i]))
  lines(res.svb[,2], res.svb[,1], type = "l", col = 2)
  lines(res.bvsnlp[,2], res.bvsnlp[,1], type = "l", col = 3)
  lines(res.r2b[,2], res.r2b[,1], type = "l", col = 4)
  
  points(res.susie[96,2], res.susie[96, 1])
  points(res.svb[96,2], res.svb[96, 1])
  points(res.bvsnlp[96,2], res.bvsnlp[96, 1])
  legend("bottomright", legend = c("susie", "survival.svb", "bvsnlp"), col = c(1,2,3), lty = 1)
}

Version	Author	Date
1c71368	yunqi yang	2024-01-29

The dots indicate PIP threshold = 0.95.

3. Results using independent X, with data from null model.

par(mfrow = c(2,3),cex.axis = 1.5)

censor_lvl = c(0, 0.2, 0.4, 0.6, 0.8, 0.99)

for (i in 1:length(censor_lvl)){
  indx = which(susie$simulate.cor_type == "independent" & susie$simulate.censor_lvl == censor_lvl[i])
  pip.susie = unlist(lapply(indx, function(x) susie$susie.pip[[x]]))
  pip.survsvb = unlist(lapply(indx, function(x) survsvb$svb.pip[[x]]))
  pip.bvsnlp = unlist(lapply(indx, function(x) bvsnlp$bvsnlp.pip[[x]]))
  pip.r2b = unlist(lapply(indx, function(x) r2b$r2b.pip[[x]]))
  is_effect = unlist(lapply(indx, function(x) susie$simulate.is_effect[[x]]))
  
  ts = seq(from = 0, to = 1, by = 0.001)
  res.susie = calculate_tpr_vs_fdr(pip.susie, is_effect, ts)
  res.svb = calculate_tpr_vs_fdr(pip.survsvb, is_effect, ts)
  res.bvsnlp = calculate_tpr_vs_fdr(pip.bvsnlp, is_effect, ts)
  res.r2b = calculate_tpr_vs_fdr(pip.r2b, is_effect, ts)
  
  
  
  plot(res.susie[,2], res.susie[,1], type = "l", xlim = c(0, 1), ylim = c(0, 1), xlab = "FDR", ylab = "Power",
       main = paste0("X independent, effect 0-3", ",censor=", censor_lvl[i]))
  lines(res.svb[,2], res.svb[,1], type = "l", col = 2)
  lines(res.bvsnlp[,2], res.bvsnlp[,1], type = "l", col = 3)
  lines(res.r2b[,2], res.r2b[,1], type = "l", col = 4)
  
  points(res.susie[96,2], res.susie[96, 1])
  points(res.svb[96,2], res.svb[96, 1])
  points(res.bvsnlp[96,2], res.bvsnlp[96, 1])
  legend("bottomright", legend = c("susie", "survival.svb", "bvsnlp"), col = c(1,2,3), lty = 1)
}

Version	Author	Date
1c71368	yunqi yang	2024-01-29