[R-sig-ME] Need an advise on bias and MSE estimation
jas ni
jasnie111 at gmail.com
Sat Jan 21 22:10:43 CET 2017
Hi guys,
I'm about to implement the comparison in my experimental work for missing
data through EM algorithm using regression.
I want to measure the bias and MSE on the parameters mean, beta1, beta2 and
sigma and my current references are from this papers (page 206):
https://pdfs.semanticscholar.org/2fa3/699c3db05cec276c3c356c921c68723fa80e.pdf
<https://www.researchgate.net/deref/https%3A%2F%2Fpdfs.semanticscholar.org%2F2fa3%2F699c3db05cec276c3c356c921c68723fa80e.pdf>
I do confuse between m and j variables.
I also refer to the attached paper and do confuse between t and j variables.
Both papers presented to measure the bias. But both equations has different
formula.
I could not implement the equation in R code because i do confuse with the
variables for instance, m and j as shown in the equation in paper 1.
I using the EM algorithm which is each iteration would produce the
parameter mean, variance and beta. But the parameter estimation values
would be generated based on the number of iteration but not based on the
number of records or data points.
If you could explain, why in the bias equation we need to sum the parameter
values up to the number of data points. For example, i have the Old
Faithful geyser dataset that contains 272 instances. Does it mean that i
should generate the 272 estimated mean, variances and betas and sum all of
them together?
- J
[[alternative HTML version deleted]]
More information about the R-sig-mixed-models
mailing list