Quantile Regression: Criteria
This feature requires Custom Tables and Advanced Statistics.
The Criteria dialog provides options for
- Quantile
- Provides options for specifying the quantile(s).
- Specify single quantiles
- When selected, at least one value is required to run the analysis. Multiple
values are allowed and each value must belong to
[0, 1]
. You can specify multiple values with each value separated by a blank space (or blank spaces). Use the Add, Change, and Remove buttons to work with the values in the quantile value list. - Specify grid quantiles
- When selected, a grid of quantiles can be specified from a
Start value (
value1
) to an End value (value2
) with the increment of By (value3
). If specified, only one valid set of [value1 TO value2 BY value3
] is allowed. It must satisfy that0 ≤ value1 ≤ value2 ≤ 1
. In cases wherevalue1 = value2
, it is equivalent to specifying a singlevalue1
, regardless ofvalue3
.
- Estimation Method
- Provides options for specifying the model estimation method.
- Automatically chosen by the program
- Allows the procedure to automatically select the appropriate estimation method. This is the default setting.
- Simplex algorithm
- Calls the simplex algorithm that was developed by Barrodale and Roberts.
- Frisch-Newton interior-point non-linear optimization
- Calls for the Frisch-Newton interior-point non-linear optimization algorithm.
- Post-estimation
- Provides options for the post-estimation of the variance-covariance of the parameter estimates
and the confidence intervals for the predicted target values.
- Assume cases are IID
- When selected, this setting assumes that error terms are independently and identically distributed. When the setting is not selected, the computation time may significantly increase for large models. The setting is selected by default.
- Bandwidth type
- Determines which bandwidth method is used to estimate the variance-covariance matrix of the parameter estimates (Bofinger or Hall-Sheather). Bofinger is the default setting.
- Numerical Method
- Provides the following options:
- Singularity tolerance
- Specifies the tolerance value for the matrix manipulations in the interior-point method. The specified value must be a single, double value in (0, 10-3), with 10-12 as the default setting.
- Convergence
- Specifies the convergence criterion for the numerical method. The specified value must be a single, double value in (0, 10-3), with 10-6 as the default setting.
- Maximum iterations
- Specifies the maximum number of iterations. The specified value must be a single, positive integer. The default value is 2000.
- Missing Values
- Provides options for determining how missing values are handled.
- Exclude both user-missing and system missing values
- When selected, both user-missing and system missing values are excluded.
- User-missing values are treated as valid
- When selected, user-missing values are treated as valid.
- Confidence interval (%)
- Specifies the significance level. When specified, the value must be a single double value in between 0 and 100. The default value is 95.