I am having particular difficulty with a deconvolution problem in R.
I have a set of weights in pounds of a product that are contaminated with measurement error. There are 72 data points. I also have a measurement standard deviation for scale used to measure the products. I want to recover an estimate of the true uncontaminated distribution.
I am assuming that the measurement errors are normally distributed and that an additive error model applies.
I am trying to use the following code to deconvolve the data:
Weight <- c(1.4760, 1.4740, 1.4700, 1.4760, 1.4720, 1.4760, 1.4760, 1.4760, 1.4746, 1.4746, 1.4710, 1.4710, 1.4730, 1.4730, 1.4710, 1.4710, 1.4705, 1.4705, 1.4756, 1.4756, 1.4770, 1.4770, 1.4752, 1.4752, 1.4735, 1.4735, 1.4743, 1.4743, 1.4747, 1.4747, 1.4736, 1.4736, 1.4732, 1.4732, 1.4731, 1.4731, 1.4752, 1.4752, 1.4760, 1.4760, 1.4746, 1.4746, 1.4709, 1.4709, 1.4754, 1.4754, 1.4705, 1.4705, 1.4707, 1.4707, 1.4708, 1.4708, 1.4753, 1.4753, 1.4692, 1.4692, 1.4754, 1.4754, 1.4700, 1.4700, 1.4752, 1.4752, 1.4764, 1.4764, 1.4748, 1.4748, 1.4749, 1.4749, 1.4745, 1.4745, 1.4671, 1.4671)
plot(density(weight))
# Producing this image of the density
[Original Data Density Plot][1]
# Now Generate the noise vector
stddev <- 0.00000606
mn <- 0
noise <- rnorm(n = 72,mean = mn, sd = stddev)
# and finally perform the deconvolution
ST <- deamerSE(y = Weight,
error = noise,
grid.length = 100)
# Then view the plotted results
plot(density(Weight))
lines(ST)
[Original plus deconvoluted density][2]
#####
I cannot get the deconvolved density to look like the original density even though the standard deviation of measurement is very small.
Any help would be appreciated.