kclass: kclass Instrumental variable estimator.
In potterzot/rkclass: K-class regressions (Two-Stage Least Squares, LIML, Fuller)

Description Usage Arguments Details Value See Also Examples

kclass fits instrumental variables models using k-class estimators.

1
2
3

kclass(formula, data, ..., weights = NULL, subset = NULL, offset = NULL,
  model = TRUE, x = FALSE, y = TRUE, z = FALSE, xz = FALSE,
  savefirst = FALSE)

`formula`	formula specification: y ~ X1 + X2 \| Z1 + Z2. If the right side is missing, the returned result is an OLS estimator.
`data`	a data frame or list containing the variables named in the formula.
`...`	additional parameters passed to kclass.fit.
`weights`	an optional vector of weights to be used in estimation.
`subset`	optional subset of observations to be used in fitting.
`offset`	known components with coefficient 1 to be included in the predictor. This should be NULL or a vector of length equal to the number of cases.
`model,x,y,z`	logicals indicating whether to return the model frame, regressors, response, and instruments used.
`savefirst`	logical if the first stage fitted values (or projected matrix) should be saved.

K-class estimators are a generic class of instrumental regression estimators. The k is a parameter that informs the calculation of coefficients. In particular, it informs the weight of the instruments. K-class estimators include two-stage least squares (TSLS) where k=1, limited information maximum likelihood (LIML) where k is the minimum eigenvalue of the LIML matrix, and FULLER in which k is the LIML k plus an additional alpha parameter. In principal, k can take on any value, though values other than those above are not generally used.

kclass returns an object of class "kclass". The function summary is used to print a summary table of results. The generic accessor functions coefficients, fitted.values, residuals, and effects provide those values.

A kclass object is a list containing the following properties:

`call`	the matched call used to create the estimators.
`coefficients`	a named vector of coefficients.
`residuals`	the residuals, calculated as the response less fitted values.
`fitted.values`	the fitted mean values. In OLS this is X'b.
`weights`	the specified weights.
`df.residual`	the residual degrees of freedom.
`terms`	the model `terms` object.
`model`	if requested, the model frame used.
`x`	if requested, the endogenous variables used.
`y`	if requested, the response variable used.
`z`	if requested, the exogenous instruments used.

AER::ivreg

library(rkclass)

#mroz
data(mroz)

# LIML model with interaction of YOB and QOB as instruments
model = kclass(LWKLYWGE ~ EDUC + as.factor(YOB) | . - EDUC + interaction(YOB, QOB), data = qob)
summary(model)
#     Estimate  Std. Error
#EDUC 0.089116   0.16110
#etc...

## Not run: 
#This is equivalent to `ivreg` from the AER package:
maer = AER::ivreg(LWKLYWGE ~ EDUC + as.factor(YOB) | . - EDUC + interaction(YOB, QOB), data = qob)
summary(maer)

#And similar to using GMM directly (warning: running this use serious resources!):
mgmm = gmm::gmm(lwage ~ exper + expersq + educ, ~ age + kidslt6 + kidsge6, data=mroz)

## End(Not run)