itersur: Seamingly Unrelated Regression (SUR) Estimation using FGLS

In hannesdatta/marketingtools: What the Package Does (One Line, Title Case)

Seamingly Unrelated Regression (SUR) Estimation using FGLS

Description

Estimates a SUR model on a possibly unbalanced panel, as layed out in Zellner (1962) and Bronnenberg, Mahajan, and Vanhonacker (2000). Supports Praise-Winsten correction for auto-correlation in the residuals.

Usage

itersur(
  X,
  Y,
  index,
  method = "FGLS",
  maxiter = 1000,
  reltol = 10^-7,
  to.file = F
)

Arguments

X	matrix of explanatory variables, possibly constructed using attraction_data.
Y	vector of dependent variable, possibly constructed using attraction_data.
index	data.frame with two columns, describing the index of the data set. The first column needs to list an indicator for time (e.g., weeks); the second column lists an indicator for cross-sectional units (e.g., factor, character, or numeric). The columns need not be named. Index can be constructed using attraction_data.
method	Choice of "FGLS" (for feasible generalized least squares), or "FGLS-Praise-Winsten" for feasible generalized least squares with Praise-Winsten transformation of the dependent and explanatory variables (to correct for autocorrelation in the residuals) (see van Heerde, Leeflang, and Wittink 2004; Greene 2003.
maxiter	Maximum number of FGLS iterations (set to 1000 by default)

Value

returns on object of class "itersur" with estimation results.

The functions \code{show} are used to obtain and print a summary of the results. The generic accessor function \code{coefficients} extracts the coefficients as a \code{data.frame}.

An object of class \code{"itersur"} is an S4-class containing at least the following components, which can be accessed using the \code{@}-operator:
\item{coefficients}{a \code{data.frame} with coefficients}
\item{X}{the predictor matrix to estimate the final iteration of the model}
\item{Y}{the dependent variable to estimate the final iteration of the model (unchanged for FGLS, transformed for Praise-Winsten.)}
\item{index}{The indexing \code{data.frame}, with column 1 being the indicators for time periods, and column 2 being the indicators for cross-sectional units.}
\item{iterations}{The number of iterations of the algorithm until convergence.}
\item{bic}{The BIC for the model.}
\item{predicted}{the fitted values}
\item{resid}{the residuals, that is response minus fitted values.}
\item{varcovar}{The variance-covariance matrix of the estimates.}
\item{method}{The method used to estimat the model; one of code{"FGLS"} or \code{"FGLS-Praise-Winsten"}.}
\item{rho_hat}{A numeric vector, indicating the estimated auto-correlation using method \code{"FGLS-Praise-Winsten"}. Zero if method \code{"FGLS"} is used.}
\item{rho}{A numeric vector, indicating the (possibly remaining) auto-correlation in the residuals, by cross-sectional unit.}

References

Bronnenberg, B. J., Mahajan, V., & Vanhonacker, W. R. (2000). The Emergence of Market Structure in New Repeat-purchase Categories: The Interplay of Market Share and Retailer Distribution. Journal of Marketing Research, 37(1), 16-31.

Greene, W. H. (2003). Econometric Analysis. Pearson Education.

van Heerde, H. J., Leeflang, P. S., & Wittink, D. R. (2004). Decomposing the Sales Promotion Bump with Store Data. Marketing Science, 23(3), 317-334.

Zellner, A. (1962). An Efficient Method of Estimating seemingly Unrelated Regressions and Tests for Aggregation Bias. Journal of the American Statistical Association, 57(298), 348-368.

Examples

# See examples listed under attraction_data.

hannesdatta/marketingtools documentation built on June 3, 2022, 11:42 p.m.