Register to reply

Estimating parameters from multivariate normal

Share this thread:
jimmy1
#1
May17-08, 11:31 AM
P: 61
I need to estimate parameters from data that follow a mutinormal distribution. The parameters that I need to estimate are contained in the expression for the mean of the marginal normal distributions. That is each marginal distribution has mean:

[tex] \frac{p_1*c_i + p_1*y}{p_1+p_2}[/tex]

where [tex]p_1[/tex], [tex]p_2[/tex] and [tex]y[/tex] are the paramters that I need to estimate and [tex]c_i[/tex] is just a known contant associated with the ith marginal. [tex]p_1[/tex] and [tex]p_2[/tex] are random parameters and [tex]y[/tex] is a fixed parameter.

I've tried using a multinormal likelihood approach, which gives an estimate for the mean as the sample mean, but how do I get estimate of the actual parameters from this sample mean? Can I even use this approach?

Also is there an added complexity to estimating parameters when there are random parameters, as is the case with [tex]p_1[/tex] and [tex]p_2[/tex]?

Any help, on how best to estimate the paramters [tex]p_1[/tex], [tex]p_2[/tex] and [tex]y[/tex] would be great?
Phys.Org News Partner Science news on Phys.org
NASA team lays plans to observe new worlds
IHEP in China has ambitions for Higgs factory
Spinach could lead to alternative energy more powerful than Popeye

Register to reply

Related Discussions
Multivariate taylor Calculus & Beyond Homework 5
Multivariate urn problem Set Theory, Logic, Probability, Statistics 0
Multivariate Exponential distribution Set Theory, Logic, Probability, Statistics 1
Multivariate Regressions? Set Theory, Logic, Probability, Statistics 13