Last updated: 2021-05-28

Checks: 7 0

Knit directory: misc/analysis/

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.


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.

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.

The command set.seed(1) 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.

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

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

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

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 e31a06c. 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:    .DS_Store
    Ignored:    .Rhistory
    Ignored:    .Rproj.user/
    Ignored:    analysis/.RData
    Ignored:    analysis/.Rhistory
    Ignored:    analysis/ALStruct_cache/
    Ignored:    data/.Rhistory
    Ignored:    data/pbmc/

Untracked files:
    Untracked:  .dropbox
    Untracked:  Icon
    Untracked:  analysis/GHstan.Rmd
    Untracked:  analysis/GTEX-cogaps.Rmd
    Untracked:  analysis/PACS.Rmd
    Untracked:  analysis/Rplot.png
    Untracked:  analysis/SPCAvRP.rmd
    Untracked:  analysis/admm_02.Rmd
    Untracked:  analysis/admm_03.Rmd
    Untracked:  analysis/compare-transformed-models.Rmd
    Untracked:  analysis/cormotif.Rmd
    Untracked:  analysis/cp_ash.Rmd
    Untracked:  analysis/eQTL.perm.rand.pdf
    Untracked:  analysis/eb_prepilot.Rmd
    Untracked:  analysis/eb_var.Rmd
    Untracked:  analysis/ebpmf1.Rmd
    Untracked:  analysis/flash_test_tree.Rmd
    Untracked:  analysis/flash_tree.Rmd
    Untracked:  analysis/ieQTL.perm.rand.pdf
    Untracked:  analysis/lasso_em_03.Rmd
    Untracked:  analysis/m6amash.Rmd
    Untracked:  analysis/mash_bhat_z.Rmd
    Untracked:  analysis/mash_ieqtl_permutations.Rmd
    Untracked:  analysis/mixsqp.Rmd
    Untracked:  analysis/mr.ash_lasso_init.Rmd
    Untracked:  analysis/mr.mash.test.Rmd
    Untracked:  analysis/mr_ash_modular.Rmd
    Untracked:  analysis/mr_ash_parameterization.Rmd
    Untracked:  analysis/mr_ash_ridge.Rmd
    Untracked:  analysis/mv_gaussian_message_passing.Rmd
    Untracked:  analysis/nejm.Rmd
    Untracked:  analysis/nmf_bg.Rmd
    Untracked:  analysis/normal_conditional_on_r2.Rmd
    Untracked:  analysis/normalize.Rmd
    Untracked:  analysis/pbmc.Rmd
    Untracked:  analysis/poisson_transform.Rmd
    Untracked:  analysis/pseudodata.Rmd
    Untracked:  analysis/qrnotes.txt
    Untracked:  analysis/ridge_iterative_02.Rmd
    Untracked:  analysis/ridge_iterative_splitting.Rmd
    Untracked:  analysis/samps/
    Untracked:  analysis/sc_bimodal.Rmd
    Untracked:  analysis/shrinkage_comparisons_changepoints.Rmd
    Untracked:  analysis/susie_en.Rmd
    Untracked:  analysis/susie_z_investigate.Rmd
    Untracked:  analysis/svd-timing.Rmd
    Untracked:  analysis/temp.RDS
    Untracked:  analysis/temp.Rmd
    Untracked:  analysis/test-figure/
    Untracked:  analysis/test.Rmd
    Untracked:  analysis/test.Rpres
    Untracked:  analysis/test.md
    Untracked:  analysis/test_qr.R
    Untracked:  analysis/test_sparse.Rmd
    Untracked:  analysis/z.txt
    Untracked:  code/multivariate_testfuncs.R
    Untracked:  code/rqb.hacked.R
    Untracked:  data/4matthew/
    Untracked:  data/4matthew2/
    Untracked:  data/E-MTAB-2805.processed.1/
    Untracked:  data/ENSG00000156738.Sim_Y2.RDS
    Untracked:  data/GDS5363_full.soft.gz
    Untracked:  data/GSE41265_allGenesTPM.txt
    Untracked:  data/Muscle_Skeletal.ACTN3.pm1Mb.RDS
    Untracked:  data/Thyroid.FMO2.pm1Mb.RDS
    Untracked:  data/bmass.HaemgenRBC2016.MAF01.Vs2.MergedDataSources.200kRanSubset.ChrBPMAFMarkerZScores.vs1.txt.gz
    Untracked:  data/bmass.HaemgenRBC2016.Vs2.NewSNPs.ZScores.hclust.vs1.txt
    Untracked:  data/bmass.HaemgenRBC2016.Vs2.PreviousSNPs.ZScores.hclust.vs1.txt
    Untracked:  data/eb_prepilot/
    Untracked:  data/finemap_data/fmo2.sim/b.txt
    Untracked:  data/finemap_data/fmo2.sim/dap_out.txt
    Untracked:  data/finemap_data/fmo2.sim/dap_out2.txt
    Untracked:  data/finemap_data/fmo2.sim/dap_out2_snp.txt
    Untracked:  data/finemap_data/fmo2.sim/dap_out_snp.txt
    Untracked:  data/finemap_data/fmo2.sim/data
    Untracked:  data/finemap_data/fmo2.sim/fmo2.sim.config
    Untracked:  data/finemap_data/fmo2.sim/fmo2.sim.k
    Untracked:  data/finemap_data/fmo2.sim/fmo2.sim.k4.config
    Untracked:  data/finemap_data/fmo2.sim/fmo2.sim.k4.snp
    Untracked:  data/finemap_data/fmo2.sim/fmo2.sim.ld
    Untracked:  data/finemap_data/fmo2.sim/fmo2.sim.snp
    Untracked:  data/finemap_data/fmo2.sim/fmo2.sim.z
    Untracked:  data/finemap_data/fmo2.sim/pos.txt
    Untracked:  data/logm.csv
    Untracked:  data/m.cd.RDS
    Untracked:  data/m.cdu.old.RDS
    Untracked:  data/m.new.cd.RDS
    Untracked:  data/m.old.cd.RDS
    Untracked:  data/mainbib.bib.old
    Untracked:  data/mat.csv
    Untracked:  data/mat.txt
    Untracked:  data/mat_new.csv
    Untracked:  data/matrix_lik.rds
    Untracked:  data/paintor_data/
    Untracked:  data/running_data_chris.csv
    Untracked:  data/running_data_matthew.csv
    Untracked:  data/temp.txt
    Untracked:  data/y.txt
    Untracked:  data/y_f.txt
    Untracked:  data/zscore_jointLCLs_m6AQTLs_susie_eQTLpruned.rds
    Untracked:  data/zscore_jointLCLs_random.rds
    Untracked:  explore_udi.R
    Untracked:  output/fit.k10.rds
    Untracked:  output/fit.varbvs.RDS
    Untracked:  output/glmnet.fit.RDS
    Untracked:  output/test.bv.txt
    Untracked:  output/test.gamma.txt
    Untracked:  output/test.hyp.txt
    Untracked:  output/test.log.txt
    Untracked:  output/test.param.txt
    Untracked:  output/test2.bv.txt
    Untracked:  output/test2.gamma.txt
    Untracked:  output/test2.hyp.txt
    Untracked:  output/test2.log.txt
    Untracked:  output/test2.param.txt
    Untracked:  output/test3.bv.txt
    Untracked:  output/test3.gamma.txt
    Untracked:  output/test3.hyp.txt
    Untracked:  output/test3.log.txt
    Untracked:  output/test3.param.txt
    Untracked:  output/test4.bv.txt
    Untracked:  output/test4.gamma.txt
    Untracked:  output/test4.hyp.txt
    Untracked:  output/test4.log.txt
    Untracked:  output/test4.param.txt
    Untracked:  output/test5.bv.txt
    Untracked:  output/test5.gamma.txt
    Untracked:  output/test5.hyp.txt
    Untracked:  output/test5.log.txt
    Untracked:  output/test5.param.txt

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/mr_ash_pen.Rmd) and HTML (docs/mr_ash_pen.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 e31a06c Matthew Stephens 2021-05-28 workflowr::wflow_publish(“analysis/mr_ash_pen.Rmd”)
html 800415b Matthew Stephens 2021-05-11 Build site.
Rmd f77f070 Matthew Stephens 2021-05-11 wflow_publish(“analysis/mr_ash_pen.Rmd”)

Introduction

The idea here is to code up mr.ash as a penalized method, compute gradients etc.

The penalized version of the mr.ash problem is (see Theorem 6 in mr.ash draft) \[\min_b (1/2\sigma^2) ||y - Xb||_2 + (1/\sigma^2) \sum_j \rho_{g,s_j}(b_j) + 0.5(n-p) \log(2\pi \sigma^2)\] Where the penalty \(\rho\) depends on the prior (\(g\)) and the \(s_j^2:=\sigma^2/(x_j'x_j)\). Here I’m going to assume for simplicity that \(g\) denotes the prior on b, ie not scale the prior on b by the residual variance as in the paper.

The penalty \(\rho_{g,s_j}(b)\) is inconvenient to compute because it involves the inverse of \(S\) (the posterior mean shrinkage function) which is not analytically available to us, at least in our current state of knowledge.

A simple idea is to rewrite the problem as: \[\min_b (1/2\sigma^2) ||y - XS_{g,s}(b)||_2 + (1/\sigma^2) \rho_{g,s}(S_{g,s}(b)) + 0.5(n-p) \log(2\pi \sigma^2)\] Here \(S_{g,s}(b)\) means apply \(S\) element-wise to the vector \(b\). There is a slight abuse of notation here as \(s_j\) may vary across \(j\). To be explicit, the \(j\)th element is \(S_{g,s}(b)\) is equal to \(S_{g,s_j}(b_j)\).

Because \(S_{g,s}\) is invertible there is no loss of generality in writing the optimization this way. Furthermore, \[h_{g,s}(b):=\rho_{g,s}(S_{g,s}(b))\] is easy to compute. Indeed \[h_{g,s}(b) = -s^2 l_{g,s}(b) - 0.5 s^4 [l_{g,s}'(b)]^2\]

where \(l_{g,s}(b)\) is the marginal log-likelihood function under the normal means model with prior \(g\) and variance \(s\). That is \(l_{g,s}(b) = \log(f(b))\) where \[f(b) := \sum_k \pi_k N(b; 0, \sigma_k^2+s^2).\]

Note: Tweedie’s formula says that relates shrinkage in the normal means model \(S\) to the marginal loglikelihood \(l\), via \[S_{f,s^2}(b) = b + s^2 l'(b)\]. So the term \(s^4 [l'(b)]^2\) above is \((S(b)-b)^2\).

Summary of optimization problem

I wanted to give a succinct summary of the overall optimization problem to share with others.

For fixed \(\sigma^2\), and letting \(w\) represent the mixture proportions in the prior \(g\) (which we often denote \(\pi\)) we can write it as follows.

Given \(y \in R^n\) and \(X \in R^{n \times p}\), \[\min ||y - XS(b; w)||_2 + h(b,w)\] subject to: \(w \in S_K\) the simplex of dimension \(K\), and \(b \in R^p\).

\[h(b,w) = -s^2 l_{w,s}(b) - 0.5 s^4 [l_{w,s}'(b)]^2\]

Code for fundamental functions

Everything can be written in terms of the marginal likelihood \(f\) and its first two derivatives, so we code those up first. I have not been careful about numerical issues here.. will need to deal with those at some point.

#y,s are vectors of length n
#w, sigma are vectors of length K (w are prior mixture proportions)
# returns an n-vector of "marginal likelihoods" under mixture prior
f = function(b, s, w, sigma){
  if(length(s)==1){s = rep(s,length(b))}
  sigmamat   <- outer(s^2, sigma^2, `+`) # n time k
  llik_mat   <- -0.5 * (log(sigmamat) + b^2 / sigmamat)
  #llik_norms <- apply(llik_mat, 1, max)
  #L_mat      <- exp(llik_mat - llik_norms)
  L_mat <- exp(llik_mat)
  return((1/sqrt(2*pi)) * as.vector(colSums(w * t(L_mat))))
}

#returns a vector of the derivative of f evaluated at each element of y
f_deriv = function(b, s, w, sigma){
  if(length(s)==1){s = rep(s,length(b))}
  sigmamat   <- outer(s^2, sigma^2, `+`) # n time k
  llik_mat   <- -(3/2) * log(sigmamat) -0.5* b^2 / sigmamat
  #llik_norms <- apply(llik_mat, 1, max)
  #L_mat      <- exp(llik_mat - llik_norms)
  L_mat <- exp(llik_mat)
  return((-b/sqrt(2*pi)) * as.vector(colSums(w * t(L_mat))))
}

# returns f_deriv/b ok even if b=0
f_deriv_over_b = function(b, s, w, sigma){
  if(length(s)==1){s = rep(s,length(b))}
  sigmamat   <- outer(s^2, sigma^2, `+`) # n time k
  llik_mat   <- -(3/2) * log(sigmamat) -0.5* b^2 / sigmamat
  #llik_norms <- apply(llik_mat, 1, max)
  #L_mat      <- exp(llik_mat - llik_norms)
  L_mat <- exp(llik_mat)
  return((-1/sqrt(2*pi)) * as.vector(colSums(w * t(L_mat))))
}

#returns a vector of the second derivatives of f evaluated at each element of y
f_deriv2 = function(b, s, w, sigma){
  if(length(s)==1){s = rep(s,length(b))}
  sigmamat   <- outer(s^2, sigma^2, `+`) # n time k
  llik_mat   <- -(5/2) * log(sigmamat) -0.5* b^2 / sigmamat
  #llik_norms <- apply(llik_mat, 1, max)
  #L_mat      <- exp(llik_mat - llik_norms)
  L_mat <- exp(llik_mat)
  return((b^2/sqrt(2*pi)) * as.vector(colSums(w * t(L_mat)))+ f_deriv_over_b(b,s,w,sigma))
}

Check the derivative code numerically:

n = 100
k = 5
b = rnorm(n)
w = rep(1/k,k)
prior_grid = c(0,1,2,3,4,5)
eps=1e-5

plot((f(b+eps,1,w,prior_grid)-f(b,1,w,prior_grid))/eps, f_deriv(b,1,w,prior_grid),xlab="numerical 1st derivative", ylab="analytic 1st derivative")
abline(a=0,b=1)

Version Author Date
800415b Matthew Stephens 2021-05-11
plot((f_deriv(b+eps,1,w,prior_grid)-f_deriv(b,1,w,prior_grid))/eps, f_deriv2(b,1,w,prior_grid), xlab="numerical 2nd derivative", ylab="analytic 2nd derivative")
abline(a=0,b=1)

Version Author Date
800415b Matthew Stephens 2021-05-11

Now we have \[l(b) = \log f(b)\], \[l'(b) = f'(b)/f(b)\], \[l''(b) = (f(b)f''(b)-f'(b)^2)/f(b)^2\],

l = function(b, s, w, prior_grid){
  log(f(b,s,w,prior_grid))
}

l_deriv = function(b, s, w, prior_grid){
  f_deriv(b,s,w,prior_grid)/f(b,s,w,prior_grid)
}

l_deriv2 = function(b, s, w, prior_grid){
  ((f_deriv2(b,s,w,prior_grid)*f(b,s,w,prior_grid))-f_deriv(b,s,w,prior_grid)^2)/f(b,s,w,prior_grid)^2
}

plot((l_deriv(b+eps,1,w,prior_grid)-l_deriv(b,1,w,prior_grid))/eps, l_deriv2(b,1,w,prior_grid), xlab="numerical 2nd derivative", ylab="analytic 2nd derivative")
abline(a=0,b=1)

Version Author Date
800415b Matthew Stephens 2021-05-11

And with these in place we can compute the shrinkage function and penalty function \(h\), using \[h(b) = -l(b) - 0.5(l'(b))^2\] \[S(b) = b+ s^2l'(b)\] \[S'(b) = 1 + s^2 l''(b)\].

S = function(b, s, w, prior_grid){
  return(b + s^2 * l_deriv(b,s,w,prior_grid))
}
S_deriv = function(b, s, w, prior_grid){
  return(1+ s^2 * l_deriv2(b,s,w,prior_grid))
}
h = function(b,s,w,prior_grid){
  return(-s^2 * l(b,s,w,prior_grid) - 0.5 * s^4 * l_deriv(b,s,w,prior_grid)^2)
}
h_deriv = function(b,s,w,prior_grid){
  return(-s^2 * l_deriv(b,s,w,prior_grid) - s^4 * l_deriv(b,s,w,prior_grid) * l_deriv2(b,s,w,prior_grid))
}


plot((h(b+eps,1,w,prior_grid)-h(b,1,w,prior_grid))/eps, h_deriv(b,1,w,prior_grid), xlab="numerical 1st derivative", ylab="analytic 1st derivative")
abline(a=0,b=1)

Version Author Date
800415b Matthew Stephens 2021-05-11

Inverting S by Newton-Raphson method

One thing we might want to do is invert S. We don’t have analytic form for this, but we can use the Newton-Raphson method to solve \(S(x)=y\). The iterates would be \[x \leftarrow x - (S(x)-y)/S'(x) \]

y = seq(-10,10,length=100)
x=y 
par(mfrow=c(5,5))
par(mar=rep(1.5,4))
w = rep(0.5,2)
prior_grid = c(0,5)
for(i in 1:25){
  plot(S(x,1,w,prior_grid),y)
  x = x-(S(x,1,w,prior_grid)-y)/S_deriv(x,1,w,prior_grid)
}

Version Author Date
800415b Matthew Stephens 2021-05-11
plot(S(x,1,w,prior_grid),y)
plot(x,y)

Version Author Date
800415b Matthew Stephens 2021-05-11

Optimizing the objective

obj = function(b, y, X, residual_var, w, prior_grid){
  d = colSums(X^2)
  s = sqrt(residual_var/d)
  r = y - X %*% S(b,s,w,prior_grid)
  return(0.5*(1/residual_var)*sum(r^2) + (1/residual_var)*sum(h(b,s,w,prior_grid)))
}

obj_grad = function(b, y, X, residual_var, w, prior_grid){
  d = colSums(X^2)
  s = sqrt(residual_var/d)
  r = y - X %*% S(b,s,w,prior_grid)
  return((-1/residual_var)*(t(r) %*% X) * S_deriv(b,s,w,prior_grid) + (1/residual_var)* h_deriv(b,s,w,prior_grid))
}

Try it out on simple simulation:

n = 100
p = 20
X = matrix(rnorm(n*p),nrow=n)
norm = colSums(X^2)
X = t(t(X)/sqrt(norm))
b = rnorm(p)
y = X %*% b + rnorm(n)

(obj(b,y,X,1,w,prior_grid)-obj(b+c(eps,rep(0,p-1)),y,X,1,w,prior_grid))/eps
[1] 0.2996978
-obj_grad(b,y,X,1,w,prior_grid)[1]
[1] 0.2997033
b.cg.warm = optim(b,obj,gr=obj_grad,method="CG",y=y,X=X,residual_var=1, w=w,prior_grid=prior_grid)
b.cg.null = optim(rep(0,p),obj,gr=obj_grad,method="CG",y=y,X=X,residual_var=1, w=w,prior_grid=prior_grid)
b.bfgs.warm = optim(b,obj,gr=obj_grad,method="BFGS",y=y,X=X,residual_var=1, w=w,prior_grid=prior_grid)
b.bfgs.null = optim(rep(0,p),obj,gr=obj_grad,method="BFGS",y=y,X=X,residual_var=1, w=w,prior_grid=prior_grid)


b.cg.warm$value
[1] 94.80382
b.cg.null$value
[1] 94.80382
b.bfgs.warm$value
[1] 94.80382
b.bfgs.null$value
[1] 94.80382
plot(S(b.cg.warm$par,1,w,prior_grid),b)

Version Author Date
800415b Matthew Stephens 2021-05-11

Some comments from Mihai: BFGS stores a dense approximation to the Hessian, so won’t be good for big problems. However, “limited”-BFGS might work (L-BFGS?). CG needs preconditioning in general; BFGS does not because it is computing an approximation to the Hessian.

Thoughts from me: maybe we can compute the Hessian directly and efficiently when X is, say, the trend filtering matrix.

Another question: if we write h as a function of pi, is h(pi) convex? Is rho(pi) convex?

Trendfiltering

Here I try a really hard trend-filtering example.

set.seed(100)
sd = 1
n = 100
p = n
X = matrix(0,nrow=n,ncol=n)
for(i in 1:n){
  X[i:n,i] = 1:(n-i+1)
}
btrue = rep(0,n)
btrue[40] = 8
btrue[41] = -8

Y = X %*% btrue + sd*rnorm(n)
norm = colSums(X^2)
X = t(t(X)/sqrt(norm))
btrue = btrue * sqrt(norm)
plot(Y)
lines(X %*% btrue)

Version Author Date
800415b Matthew Stephens 2021-05-11
prior_grid=c(0,10,100,1000)
w=rep(1/4,4)
b.cg.null = optim(rep(0,p),obj,gr=obj_grad,method="CG",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid,control=list(maxit=1000))
b.bfgs.null = optim(rep(0,p),obj,gr=obj_grad,method="BFGS",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid,control=list(maxit=1000))

plot(Y)
lines(X %*% S(b.cg.null$par,1,w,prior_grid))
lines(X %*%  S(b.bfgs.null$par,1,w,prior_grid))


b.cg.warm = optim(btrue,obj,gr=obj_grad,method="CG",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid)
b.bfgs.warm = optim(btrue,obj,gr=obj_grad,method="BFGS",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid)
lines(X %*% S(b.cg.warm$par,1,w,prior_grid))
lines(X %*% S(b.bfgs.warm$par,1,w,prior_grid),col=2)

Version Author Date
800415b Matthew Stephens 2021-05-11
b.cg.warm$value
[1] 289.8884
b.cg.null$value
[1] 385.2665
b.bfgs.warm$value
[1] 289.8891
b.bfgs.null$value
[1] 422.1747

This case is ridge regression, so should be convex… try it out.

prior_grid=c(10,10)
w=rep(1/2,2)
b.cg.null = optim(rep(0,p),obj,gr=obj_grad,method="CG",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid,control=list(maxit=1000))
b.bfgs.null = optim(rep(0,p),obj,gr=obj_grad,method="BFGS",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid,control=list(maxit=1000))

plot(Y)
lines(X %*% S(b.cg.null$par,1,w,prior_grid))
lines(X %*%  S(b.bfgs.null$par,1,w,prior_grid))

Version Author Date
800415b Matthew Stephens 2021-05-11
b.bfgs.null$value
[1] 428.7797
b.cg.null$value
[1] 428.7797

Try a prior with more overlapping components

prior_grid=seq(0,100,length=100)
w=rep(1/100,100)
b.cg.null = optim(rep(0,p),obj,gr=obj_grad,method="CG",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid,control=list(maxit=1000))
b.bfgs.null = optim(rep(0,p),obj,gr=obj_grad,method="BFGS",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid,control=list(maxit=1000))

plot(Y)
lines(X %*% S(b.cg.null$par,1,w,prior_grid))
lines(X %*%  S(b.bfgs.null$par,1,w,prior_grid))

Version Author Date
800415b Matthew Stephens 2021-05-11
b.bfgs.null$value
[1] 462.9663
b.cg.null$value
[1] 472.9022

And now revert to the sparser prior

prior_grid=c(0,10,100,1000)
w=rep(1/4,4)
b.cg.null = optim(b.cg.null$par,obj,gr=obj_grad,method="CG",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid,control=list(maxit=1000))
b.bfgs.null = optim(b.bfgs.null$par,obj,gr=obj_grad,method="BFGS",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid,control=list(maxit=1000))
plot(Y)
lines(X %*% S(b.cg.null$par,1,w,prior_grid))
lines(X %*%  S(b.bfgs.null$par,1,w,prior_grid))

Version Author Date
800415b Matthew Stephens 2021-05-11
b.bfgs.null$value
[1] 302.7868
b.cg.null$value
[1] 310.1688

Try initializing from bhat the “overfit” solution.

bhat = chol2inv(chol(t(X) %*% X)) %*% t(X) %*% Y
plot(X %*% bhat)
lines(Y)

Version Author Date
800415b Matthew Stephens 2021-05-11
b.bfgs.bhat = optim(bhat,obj,gr=obj_grad,method="BFGS",y=Y,X=X,residual_var=1, w=w,prior_grid=prior_grid,control=list(maxit=10000))
b.bfgs.bhat$value
[1] 293.4889
plot(Y)
lines(X %*%  S(as.vector(b.bfgs.null$par),1,w,prior_grid))
lines(X %*%  S(as.vector(b.bfgs.warm$par),1,w,prior_grid),col=2)
lines(X %*%  S(as.vector(b.bfgs.bhat$par),1,w,prior_grid),col=3)

Version Author Date
800415b Matthew Stephens 2021-05-11

sessionInfo()
R version 3.6.0 (2019-04-26)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS  10.16

Matrix products: default
BLAS:   /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.6/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

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

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.6       rstudioapi_0.13  whisker_0.4      knitr_1.29      
 [5] magrittr_2.0.1   workflowr_1.6.2  R6_2.4.1         rlang_0.4.10    
 [9] stringr_1.4.0    tools_3.6.0      xfun_0.16        git2r_0.27.1    
[13] htmltools_0.5.0  ellipsis_0.3.1   yaml_2.2.1       digest_0.6.27   
[17] rprojroot_1.3-2  tibble_3.0.4     lifecycle_1.0.0  crayon_1.3.4    
[21] later_1.1.0.1    vctrs_0.3.8      fs_1.5.0         promises_1.1.1  
[25] glue_1.4.2       evaluate_0.14    rmarkdown_2.3    stringi_1.4.6   
[29] compiler_3.6.0   pillar_1.4.6     backports_1.1.10 httpuv_1.5.4    
[33] pkgconfig_2.0.3