makefunobj |

Export functional objects to matrices in the workfile.

For equations estimated using the functional coefficients method.

Syntax

eq_name.makefunobj(options) matrix_name

where matrix_name is the name of the output matrix.

Options

Basic Options

type=arg (default= “coef”) | Type of result to save: coefficients (“coef”), residuals (“res”), bias (“bias”), covariances involving a single coefficient (“cov”), correlations involving a single coefficient (“cor”), confidence intervals (“ci”). If saving covariances or correlations, you may identify the coefficient of interest using the “coefid=” option. |

wf | Make output of workfile length. Note: this option is only relevant if estimation was evaluated over the functional coefficient variable values. |

derivs | Include derivative coefficients as part of the output (for all but “type=resid”). Use derivative coefficients are part of the calculation (when type=resid”). |

nodups | Duplicate observations for in the set of functional coefficient evaluation points are removed. |

sort | Rows of the output matrix are sorted in increasing order of the functional coefficient evaluation points. |

Pilot Bandwidth Options

If a local pilot bandwidth has exists, all of the pilot bandwidth computation options below will be ignored unless the “nolocalbw” option is specified.

plth =arg (default = “cv”) | Pilot bandwidth method: simple rule-of-thumb (“rot”), robust rule-of-thumb (“rotr”), residual squares criterion (“rsc”), modified multi cross-validation (“cv”), user-defined (“user”). |

pltbw=arg (default =1) | User-defined bandwidth (if “plth=user”). |

plthmin=arg (default = 0.1) | Bandwidth grid search minimum value (if not “plth=user”). |

plthmax=arg (default =1) | Bandwidth grid search maximum value (if not “plth=user”). |

plthlen=integer (default = 100) | Bandwidth grid search length (if not “plth=user”). |

plthinc=integer (default = 10) | Bandwidth grid search increment step percentage increase (if not “plth=user”). |

plthcup=integer (default = 10) | Stop rule: consecutive increases of objective function before stop (not available when “plth=user”). |

pltm=arg (default = 10) | Modified multifold CV m-value: percentage of sample size used in bandwidth determination (when “plth=cv”). |

pltq=integer (default = 4) | Modified multifold CV Q-value: percentage of sample size used in bandwidth determination (when “plth=cv”). |

auxk=integer (default = 2) | Estimation polynomial degree for pilot stage in excess of final stage degree. This number should always be an even positive integer. |

Covariance Options

coefid=integer (default = 1) | Coefficient ID for which covariances or correlations are produced when “type=cov” or “type=cov”. |

Confidence Interval Options

The following options are only available when “type=ci”.

seband | Produce standard error bands instead of confidence intervals. |

sewidth =integer (default = 1) | Number of standard errors to use as half-width of confidence band (when “seband” is specified). |

cilevel =arg (default = 0.95) | Confidence interval coverage as a number between 0 and 1 (not applicable when “seband” is specified). |

nobias | Ignore bias in confidence interval determination. |

Examples

eq.makefunobj funcoef

produces a matrix of functional coefficients (one column per coefficient) called “FUNCOEF” using the existing local pilot bandwidth if present, or estimating the pilot bandwidth, if not. If estimated, the local pilot bandwidth is updated with the result.

eq.makefunobj(type=cor, coefid=3) myfuncorr

produces a matrix of functional correlations with respect to the third functional coefficient. The matrix is stored in the workfile under the name “MYFUNCORR”.

eq.makefunobj(nolocalbw, type=ci, cilevel=0.9) myfunci

produces a matrix with the 90% functional confidence intervals. The lower and upper bounds of the intervals are paired in adjacent columns.

Cross-references

See
“Functional Coefficient Regression” for discussion of functional coefficients estimation. See
“Post-Estimation Views and Procs” for detail on the various functional calculations.