Scribd
Upload
22 views2 pages

Iteration & Distribution: Want To Choose As Follows Argmin

The document discusses iterative methods for estimating parameters in nonlinear regression models. It describes: 1) An iterative method where the estimate from the previous iteration, θbj, is used to compute the next estimate, θbj+1, until convergence is reached. 2) The approximate distribution of the parameter estimates θb is normally distributed around the true value θ with variance proportional to the inverse of the design matrix. 3) Two methods for obtaining confidence intervals for the parameters when restrictions like non-negativity are placed on them: inverting an F-test using libraries like MASS, or using a t-statistic and threshold based on the t-distribution.

Uploaded by

Bao Gan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
22 views2 pages

Iteration & Distribution: Want To Choose As Follows Argmin

The document discusses iterative methods for estimating parameters in nonlinear regression models. It describes: 1) An iterative method where the estimate from the previous iteration, θbj, is used to compute the next estimate, θbj+1, until convergence is reached. 2) The approximate distribution of the parameter estimates θb is normally distributed around the true value θ with variance proportional to the inverse of the design matrix. 3) Two methods for obtaining confidence intervals for the parameters when restrictions like non-negativity are placed on them: inverting an F-test using libraries like MASS, or using a t-statistic and threshold based on the t-distribution.

Uploaded by

Bao Gan
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Iteration & Distribution

Want to choose θb1 as follows


● Today’s class
● Nonlinear regression models
■
● Weight loss data
● What to do? n
X
θb1 = argmin (Yi − ωi0 − Z 0 θ)2
● Delta method
● Nonlinear regression
● Nonlinear regression: details θ i=1
● Iteration & Distribution
● Confidence intervals
● Weight loss data ■ Solution
−1
θb1 = Z 0 t Z 0 Z 0 t (Y − ω 0 ).
■ Given θbj , set
−1
θbj+1 = Z j t Z j Z j t (Y − ω j ).
■ Repeat until convergence.
■ b
Approximate distribution of θ:

θb ∼ N (θ, σ bt Z)
b 2 (Z b −1 ),

where
n
X
b2 =
σ b 2 /(n − p).
(Yi − f (xi , θ)) - p. 9/11
Confidence intervals

● Today’s class
● Nonlinear regression models
■ If θ is restricted, say θ ≥ 0 the asymptotic confidence
● Weight loss data
● What to do?
intervals may be inaccurate (may overlap into negative
● Delta method numbers).
● Nonlinear regression
● Nonlinear regression: details
● Iteration & Distribution
■ library(MASS) provides another method to obtain
● Confidence intervals confidence intervals based on “inverting” an F -test.
● Weight loss data

■ Basic idea, confidence interval for θ1 : for each fixed value


θ1,0 we could compute the “extra sum of squares” between
the unrestricted model and the model with θ1 fixed at θ1,0 .

SSEθ1 =θ1,0 (θb2:p ) − SSE(θ)


b
F (θ1,0 ) = 2
∼ F1,n−p
σ
b
at least approximately under H0 : θ1 = θ1,0 .
■ Or, q
T (θ1,0 ) = sign(θb − θb1,0 ) F (θ1,0 ) ∼ tn−p .
■ Confidence interval:
{θ1 : −tn−p,1−α/2 < T (θ1 ) < tn−p,1−α/2 }. - p. 10/11

You might also like