Unrestricted GMM estimation

Click in any of the following fields (apart from the Estimation and Testing, since they are not described here) to go directly to the description of that filed.

Fields' explanation

Dataset

You must input one of the following:
A Matlab' variable that consists of the dataset and already exists in the workspace, or
A valid Matlab expression, for example randn(150,2).

The size of the data must be the one we defined in the description of the sample moments.

Moments

In this field you enter the name of the function that calculates the moment conditions and their gradient. This file must exist in Matlab's path.

For more information on the type of inputs/output of the moments' function refer to the sample moments help file.

Various arguments

If the sample moments DO NOT require instruments: Enter N/A in this field.
If the sample moments DO require the use of instuments you must input one of the following:
A Matlab' variable that consists of the instruments and already exists in the workspace, or
A valid Matlab expression, for example randn(150,5).

The size of the instruments must follow the one we defined in the sample moments help file.

Starting values

A vector with an initial guess for θ. Similarly with the fields defined thus far:
It must be a variable that exists in the workspace, or
A valid Matlab expression

Its size must be Px1.

1st step W

The weighting matrix used to calculate the first step GMM estimates.
It must exist in the workspace, or be a valid Matlab expression.

Select covariance matrix

Select the type of the moments' long-run covariance matrix using this "drop-down" menu. The methods currently supported are "Serially Uncorrelated", and "HAC" with Bartlett or Parzen weights.

Bandwidth

A non-negative integer, used for the calculation of the Bartlett or Parzen kernel.
If you right click on the edit-box, the word "Optimal" will appear; in this case, the bandwidth will be automatically calculated using Newey's method of Bandwidth selection.
If the method you selected for the moments' long run variance (previous step) is Serially Uncorrelated, a bandwidth of 0 will be used irrespectively of what you provide.

GMM iterations

A non-negative integer, indicating an upper bound for the number of the iterative GMM estimates. [Default value is 50]

Tolerance for GMM iterations

A non-negative value that controls when the search routine will stop. Specifically, if the GMM estimates on the i^thiteration are denoted by θ(i), the optimization will stop when: || θ(i) - θ(i-1)|| < Tolerance. If this condition is not satisfied, the algorithm will stop when it reaches the number of maximum GMM iteration defined above. [Default value is 1e-6]

Tolerance criterion for minimization

Used by Matlab's minimization procedure. This is the tolerance criterion for the termination of the minimization procedure (within each GMM iteration).
It must be a positive value. [Default value is 1e-6]

Max. fun evaluations

Used by Matlab's minimization procedure. Maximum number of function evaluations during the minimization procedure (within each GMM iteration).
This must be a positive integer. [Default value is 400]

Max. fun iterations

Used by Matlab's minimization procedure. Maximum number of function iterations during the minimization procedure (within each GMM iteration).
This must be a positive integer. [Default value is 400]

Display

Used by Matlab's minimization procedure. If you set it to 'off' no output is displayed; 'iter' displays output at each iteration; 'final' displays only the final output ; 'notify' dislays output only if the function does not converge.

Diagnostics

Used by Matlab's minimization procedure. Print diagnostic information about the GMM objetive function.

Derivative Check

Used by Matlab's minimization procedure. Compares user-supplied analytic derivatives to finite differencing derivatives.

Output will be saved as

The estimation output will be saved in the workspace in a "structure" variable, with the name you declare in this box (follow this link for more details concerning the output).

Estimate

Perform the estimation.

Clear all

Clears all fields and returns the GUI to its initial state.