Controlling Error Correlations in Models with Panel Data

Share:

Description

errorcorr is an additional function in the bdynsys package. It calls functions preprocess_data. It computes the errors in the models (differential equations) and the covariances of the errors. It then uses the covarinaces to re-estimate the Betas of the models. The covarinace matrix is printed and the re-estimated Betas saved in a file. It requires the package MASS

Usage

1
errorcorr(dataset, indnr, x, y, f, xterms, yterms, nrterms, z, zterms, v, vterms)

Arguments

dataset

a plm pdata.frame panel data frame.

indnr

an integer number indicating number of indicators, to be included in the modeling procedure

x

a reference to variable from the paneldata to be included as indicator 1 in the modeling procedure.

y

a reference to variable from the paneldata to be included as indicator 2 in the modeling procedure.

f

a function that contains the models of the indicators.

xterms

a vector that contains the terms from the model dx/dt.

yterms

a vector that contains the terms from the model dy/dt.

nrterms

total number of in all equations, e.g. sum of terms in equation for dx/dt and terms in equation for dy/dt, if the number of variables is two.

z

a reference to variable from the paneldata to be included as indicator 3 in the modeling procedure.

zterms

a vector that contains the terms from the model dz/dt.

v

a reference to variable from the paneldata to be included as indicator 4 in the modeling procedure.

vterms

a vector that contains the terms from the model dv/dt.

Author(s)

Viktoria Spaiser: viktoria.sp@web.de

Examples

1
2
3
4
5
## Controlling Error Correlations with two indicators and with the following two models:
## dx/dt  = + 0.0012 /x^2 and dy/dt = + 0.0071 x^3

errorcorr(datap, 2, datap$logGDP, datap$EmanzV, 
f <- function(Y=c()) rbind(0.0012/Y[1]^2, + 0.0071*Y[1]^3), c(11), c(14), 2)