'================================================================================================== '================================================================================================== '================================================================================================== subroutine local bfl (series y_lf_ser, series y_hf_ser, scalar ta, scalar dd, scalar s) ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' PURPOSE: Temporal disaggregation using Boot-Feibes-Lisman method ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' SYNTAX: series y_hf_ser ' call bfl (y_lf_ser, y_hf_ser, ta, dd, s) ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' OUTPUT: y_hf_ser high frequency estimate series name ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' INPUT: y_lf_ser low frequency series name, series must be disaggregated ' equaly on subperiods, there are not allowed NA inside sample ' ta: type of temporal disaggregation ' ta=1 sum (flow) ' ta=2 average (index) ' ta=3 last element (stock) - interpolation ' ta=4 first element (stock) - interpolation ' dd: order of diferentiation (objective function to be minimized): volatility of ... ' dd=0 levels ' dd=1 first order diferentiation ' dd=2 second order diferentiation ' s: number of high frequency points ' for each low frequency data points (freq. conversion) ' s=4 annual to quarterly ' s=12 annual to monthly ' s=3 quarterly to monthly ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' LIBRARY: d:\Eviews_library\agg_help.prg: AGG() DIF() ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' writen by: ' Slawomir Dudek ' MOF ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' REFERENCE: 1) Quilis, E.M (2002) "A MATLAB library of temporal disaggregation methods: summary" ' Instituto Nacional de Estadistica, Spain ' 2) Boot, J.C.G., Feibes, W., Lisma, J.H.C. (1967) ' "Further methods of Derivation of Quarterly Figures from Annual data" ' Applied Statistics, vol. 16, n. 1, p. 65-75. ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' Asiggning series to vectors, only not NA observations sample NA_smpl_y @all if y_lf_ser <> NA stomna(y_lf_ser, y_lf_temp, NA_smpl_y) ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' Size of the problem !n_hf=@rows(y_lf_temp) !n_lf=!n_hf/s ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' Aggregation of low frequency series equaly distributed in subperiods matrix AGG1 call aggreg (AGG1, 1, !n_lf, s, 0, 0) matrix y_lf=AGG1*y_lf_temp delete AGG1 ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' Generation of aggregation matrix AGG for procedure matrix AGG call aggreg (AGG, ta, !n_lf, s, 0, 0) ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' Generation of (implicit) VCV martix H: n x n ' depending on the selected objective function ' Generation of penalty matrix DIF n-1 x n (modified) matrix DIF call dif(DIF, dd, !n_hf, 1) matrix H=@transpose(DIF)*DIF ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' Computing high frequency series: ' 1) Generation of supermatrix AA ' 2) Expanding low frequency input ' 3) y_hf=inv(AA)*y_lf_e matrix(!n_lf, !n_lf) zeros=0 matrix(!n_hf,1) y_lf_e_zeros=0 matrix(!n_lf+!n_hf, !n_lf+!n_hf) AA matrix(!n_lf+!n_hf,1) y_lf_e matplace(AA, H, 1, 1) matplace(AA, @transpose(AGG), 1, !n_hf+1) matplace(AA, AGG, !n_hf+1, 1) matplace(AA, zeros, !n_hf+1, !n_hf+1) matplace(y_lf_e,y_lf_e_zeros,1,1) matplace(y_lf_e, y_lf, !n_hf+1, 1) matrix y_hf_e=@inverse(AA)*y_lf_e matrix y_hf = @subextract(y_hf_e,1,1,!n_hf,1) ' --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ' Assigning vector to series mtos(y_hf, y_hf_ser, NA_smpl_y) smpl @all endsub '================================================================================================== '================================================================================================== '==================================================================================================