build.hf.lm: Hedonic Function Based on a Linear Model
In hepi: Functions for estimating hedonic elementary price indices

Description Usage Arguments Details Value Author(s) References See Also Examples

View source: R/hepi.hf.R

This function estimates hedonic functions based on a linear regression model for a given dataset.

build.hf.lm(learndata, full.formula, min.formula, 
            backtrans = I, rm.infl = TRUE, 
            description = NULL, return.row.labels = FALSE, 
            allow.variable.selection = TRUE)

`learndata`	A `data.frame` containing the training data set.
`full.formula`	The formula of the full linear model. See Details.
`min.formula`	If variable selection is wanted, the formula of the minimal linear model. See Details.
`backtrans`	A backtransformation function applied to all predictions. See Details.
`rm.infl`	A logical value indicating whether influential observations should be removed.
`description`	A character string describing the hedonic function.
`return.row.labels`	A logical value indicating whether the row labels of the cleaned training data should be returned.
`allow.variable.selection`	A logical value indicating whether variable selection should be carried out.

This function estimates a hedonic function based on a linear regression model. An appropriate model formula must be given in full.formula. (See lm for more details about specifying formulae.)

The function given in backtrans is used to backtransform any predicted value using the linear model and defauls to the identity function I. If, for example, log(price) stands on the left-hand side of the model formula, any predicted value needs to be transformed with the exponential function to a valid price. This can be accomplished by indicating backtrans = exp.

If rm.infl is TRUE, influential observations having

|DFFITS_i| > 2 * sqrt(K/N)

with K being the number of exogenous variables and N the dimension of the learning data set are removed before fitting the final model. There, the DFFITS_i values are calculated based on the residuals of a first fit of a linear model using the model formula full.formula.

If allow.variable.selection is TRUE, a stepwise model selection based on exact AIC is carried out (see stepAIC for more details). In this case, full.formula acts as upper and min.formula as lower limit of the search algorithm. If allow.variable.selection is FALSE the hedonic function is estimated using exactly the formula given in full.formula.

In description, a character string describing the hedonic function may be given which is saved within the returned "hedonic.function" object.

If return.row.labels == FALSE, the function returns a "hedonic.function" object representing the fitted regression model.

If return.row.labels == TRUE, the function returns a list with following elements:

`hf`	The resulting `"hedonic.function"` object.
`row.labels`	A vector containing the row labels of the cleaned training data set.

Michael Beer r-hepi@michael.beer.name

Beer, M. (2007) Hedonic Elementary Price Indices: Axiomatic Foundation and Estimation Techniques. PhD thesis, University of Fribourg Switzerland, http://www.michael.beer.name/phdthesis.

build.hf.lm.split

data(boston, package = "spdep")

hf0 <- build.hf.lm(
    learndata = boston.c, 
    full.formula = log(MEDV) ~ CRIM + ZN + INDUS + CHAS + 
      I(NOX^2) + I(RM^2) + AGE + log(DIS) + log(RAD) + TAX + 
      PTRATIO + B + log(LSTAT), 
    backtrans = exp, 
    rm.infl =  FALSE, 
    description = NULL, 
    return.row.labels = FALSE, 
    allow.variable.selection = FALSE)

is.applicable.hf(hf0, boston.c)
summary(hf0(boston.c))

plot(boston.c$MEDV, hf0(boston.c), xlab = "Observed", ylab = "Predicted")
abline(0,1)

hf1 <- build.hf.lm(
    learndata = boston.c, 
    full.formula = log(MEDV) ~ CRIM + ZN + INDUS + CHAS + 
      I(NOX^2) + I(RM^2) + AGE + log(DIS) + log(RAD) + TAX + 
      PTRATIO + B + log(LSTAT), 
    min.formula = log(MEDV) ~ 1, 
    backtrans = exp, 
    rm.infl =  FALSE, 
    description = NULL, 
    return.row.labels = FALSE, 
    allow.variable.selection = TRUE)
summary(hf1(boston.c))