[Y2,B,RHO,U1,U2] = chowlin(Y1,X2)
[Y2,B,RHO,U1,U2] = chowlin(Y1,X2,range,...)
Y1
[ tseries ] - Low-frequency input tseries object that will be distributed over higher-frequency observations.
X2
[ tseries ] - Tseries object with regressors used to distribute the input data.
range
[ numeric ] - Low-frequency date range on which the distribution will be computed.
Y2
[ tseries ] - Output data distributed with higher frequency.
B
[ numeric ] - Vector of regression coefficients.
RHO
[ numeric ] - Actually used autocorrelation coefficient in the residuals.
U1
[ tseries ] - Low-frequency regression residuals.
U2
[ tseries ] - Higher-frequency regression residuals.
'constant='
[ true
| false
] - Include a constant term in the regression.
'log='
[ true
| false
] - Logarithmise the data before distribution, de-logarithmise afterwards.
'ngrid='
[ numeric | 200
] - Number of grid search points for finding autocorrelation coefficient for higher-frequency residuals.
'rho='
[ 'estimate'
| 'positive'
| 'negative'
| numeric ] - How to determine the autocorrelation coefficient for higher-frequency residuals.
'timeTrend='
[ true
| false
] - Include a time trend in the regression.
Chow,G.C., and A.Lin (1971). Best Linear Unbiased Interpolation, Distribution and Extrapolation of Time Series by Related Times Series. Review of Economics and Statistics, 53, pp. 372-75.
See also Appendix 2 in Robertson, J.C., and E.W.Tallman (1999). Vector Autoregressions: Forecasting and Reality. FRB Atlanta Economic Review, 1st Quarter 1999, pp.4-17.