fgls#

Purpose#

Computes a feasible generalized least squares (FGLS) regression.

Format#

out = fgls(data, formula[, method, ctl])#

out = fgls(depvar, indvars[, method, ctl])

Parameters:

data (Dataframe) – Name of dataframe.
formula (String) –
Formula string of the model. E.g "y ~ X1 + X2", y is the name of dependent variable, X1 and X2 are names of independent variables;

E.g "y ~ .", . means including all variables except dependent variable y;

E.g "y ~ -1 + X1 + X2", -1 means no intercept model.
depvar (Nx1 dataframe or matrix) – Dependent variable.
indvars (NxK matrix or dataframe) – Independent variable(s).

method (String) –

Optional input, model for innovations covariance estimate, \(\hat{\Omega}\).

`"CLM"`	\(\omega_i = \frac{1}{df}\sum_{i=1}^T \epsilon_i^2\)
`"AR"`	\(AR(1)\) model. (Default)
`"HC0"`	\(\omega_i = \epsilon_i^2\)
`"HC1"`	\(\omega_i = \frac{T}{df} \epsilon_i^2\)
`"HC2"`	\(\omega_i = \frac{\epsilon_i^2}{1 - h_i}\)
`"HC3"`	\(\omega_i = \frac{\epsilon_i^2}{1 - h_i}^2\)
`"HC4"`	\(\omega_i = \frac{\epsilon_i^2}{1 - h_i}^{d_i}\)

ctl (Struct) –

Optional input, instance of an fglsControl structure containing the following members:

ctl.intercept	Scalar, default 1. 1: a constant term will be added. 0: no constant term will be added.
ctl.iters	Scalar, number of iterations, default is two-stage FGLS.
ctl.omega0	Matrix or vector, user-defined \(\hat{\Omega}\), specified as a positive vector, positive semidefinite matrix, or a positive definite matrix. If provided, method is ignored and no data-driven \(\hat{\omega}\) is computed.
ctl.miss	Scalar, default 0. 0: Do not drop missing values. Will error if any missing values are detected. 1: listwise deletion, drop any cases in which missings occur. Generates warning message if missing values detected.
ctl.scaleResiduals	Scalar, default 0. 0: do not scale residuals. 1: Scale residuals to truncate values at the first and ninety-ninth percentiles.
ctl.mleAR	Scalar, default 1. 0: Use OLS estimate of autoregressive parameters in innovation covariance matrix estimation. 1: Use MLE estimate of autoregressive parameters in innovation covariance matrix estimation. Valid only if using `"AR"` method for computing innovation covariance matrix.
ctl.printOutput	Scalar, default 1. 0: do not print statistics. 1: print the statistics.

Returns:

out (Struct) –

instance of fglsOut struct containing the following members:

out.beta_fgls	Dx1 vector, the feasible least squares estimates of parameters.
out.sigma_fgls	DxD matrix, covariance matrix of estimated parameters.
out.se_fgls	Dx1 vector, the standard errors of the estimated parameters.
out.ci	Dx2 vector, the confidence interval of the estimated parameters.
out.t_stats	Dx1 vector, the t-statistics of the estimated parameters.
out.pvt	Dx1 vector, the p-value the t-statistics of the estimated parameters.
out.resid	Residuals, \(out.resid = y - x * out.beta\_{fgls}\).
out.df	Scalar, degrees of freedom.
out.sse	Scalar, sum of squared errors.
out.sst	Scalar, total sum of squares.
out.std_est	Scalar, standard deviation of residuals.
out.fstat	Scalar, model F-stat.
out.pvf	Scalar, p-value of model F-stat.
out.rsq	Scalar, R squared, coefficient of determination.
out.rbsq	Scalar, Rbar squared, coefficient of determination.
out.dw	Scalar, Durbin-Watson statistic.

Examples#

Basic usage with matrices#

rndseed 907808;

// Generate y matrix
y = rndn(50, 1);

//  Generate x matrix
x = rndn(50, 3);

// Perform fgls regression and print report to the screen
call fgls(y, x);

The output for data matrices includes default variable names:

Valid cases:                     50              Dependent variable:             Y
Total SS:                    48.078              Degrees of freedom:            46
R-squared:                    0.019              Rbar-squared:              -0.045
Residual SS:                 47.145              Std error of est:           1.012
F(3,46)                       0.303              Probability of F:           0.874
Durbin-Watson                 2.087


----------------------------------------------------------------------------------
                        Standard                    Prob
  Variable   Estimates       Error     t-value        >|t|  [95% Conf.   Interval]
----------------------------------------------------------------------------------

Constant     -0.0699       0.146      -0.478       0.635      -0.356       0.217
      X1       0.103       0.139       0.744       0.461      -0.169       0.376
      X2       0.166       0.183       0.906       0.370      -0.193       0.524
      X3     -0.0228       0.129      -0.177       0.860      -0.275       0.229

Basic usage with a dataframe and a formula string#

// Create string with the name and full file path of the dataset
dataset = getGAUSSHome("examples/df_returns.gdat");

// Load dataset into dataframe
data_df = loadd(dataset);

// Create formula string specifying dependent and independent variables
formula  = "rcoe ~ rcpi";

// Perform estimation
call fgls(dataset, formula);

In this example, the dataset df_returns.gdat is used to compute a regression. The dependent variable is rcoe. The independent variable is rcpi. The output is:

Valid cases:                    248              Dependent variable:            rcpi
Total SS:                     0.027              Degrees of freedom:             246
R-squared:                    0.110              Rbar-squared:                 0.107
Residual SS:                  0.024              Std error of est:             0.010
F(1,245)                     30.453              Probability of F:             0.000
Durbin-Watson                 0.757


------------------------------------------------------------------------------------
                        Standard                    Prob
Variable   Estimates       Error     t-value        >|t|  [95% Conf.   Interval]
------------------------------------------------------------------------------------

Constant      0.0148     0.00122        12.1       0.000      0.0124      0.0172
    rcoe       0.196      0.0685        2.86       0.005      0.0619        0.33

Changing method of innovation covariance estimation and storing results#

new;

// Declare 'fgls_out' to be an fglsOut structure
// to hold the results of the computations
struct fglsOut fgls_out;

// Create string with the name and full file path of the dataset
dataset = getGAUSSHome("examples/credit.dat");

// Load dataframe
data_df = loadd(dataset);

// Create a string with the name of the dependent variable
formula = "Limit ~ Balance + Income + Age";

// Specify method to be classic linear model
method = "CLM";

// Perform estimation, using CLM innovation covariance
// and store the results in 'fgls_out'
fgls_out = fgls(data_df, formula, method);

In this example, the dataset credit.dat is used to compute a regression using a classic linear model innovation covariance matrix. The dependent variable is Limit. The independent variables are: Balance, Income, and Age.

Valid cases:                    400              Dependent variable:         Balance
Total SS:            2125784986.000              Degrees of freedom:             396
R-squared:                    0.939              Rbar-squared:                 0.939
Residual SS:          129727134.947              Std error of est:           572.358
F(3,396)                   2031.029              Probability of F:             0.000
Durbin-Watson                 1.953


------------------------------------------------------------------------------------
                         Standard                    Prob
 Variable   Estimates       Error     t-value        >|t|  [95% Conf.   Interval]
------------------------------------------------------------------------------------

 Constant    1.52e+03         102        14.9       0.000    1.32e+03    1.72e+03
   Income        3.17      0.0706        44.9       0.000        3.03        3.31
      Age        32.6       0.936        34.8       0.000        30.7        34.4
    Limit        1.68        1.69        0.99       0.323       -1.64           5

Source#

fgls.src