## Problem forecasting with PDL equation

For technical questions regarding estimation of single equations, systems, VARs, Factor analysis and State Space Models in EViews. General econometric questions and advice should go in the Econometric Discussions forum.

Moderators: EViews Gareth, EViews Moderator

ghammond
Posts: 1
Joined: Fri Jun 08, 2018 10:32 am

### Problem forecasting with PDL equation

I'm experiencing a problem forecasting with an equation estimated with the PDL function. The estimated equation looks fine,

Dependent Variable: YTXY_CORPCB
Method: ARMA Maximum Likelihood (BFGS)
Date: 06/05/18 Time: 10:29
Sample: 1986Q1 2016Q1
Included observations: 121
Convergence achieved after 5 iterations
Coefficient covariance computed using outer product of gradients

Variable Coefficient Std. Error t-Statistic Prob.

C 1.27E+09 4.10E+08 3.106774 0.0024
@MOVAV(YTX_CORPSLSALLOC(-2),4) -38878448 7039342. -5.523023 0.0000
@SEAS(2) 1.71E+09 2.31E+08 7.378722 0.0000
@SEAS(3) 7.49E+08 3.02E+08 2.476446 0.0148
@SEAS(4) 4.02E+08 2.44E+08 1.646133 0.1026
PDL01 65684.73 70829.06 0.927370 0.3558
PDL02 41444.35 70536.43 0.587560 0.5580
PDL03 19443.74 11250.47 1.728260 0.0867
PDL04 -13005.23 6859.295 -1.896001 0.0606
AR(4) 0.621461 0.083447 7.447401 0.0000
SIGMASQ 1.22E+17 1.60E+16 7.632350 0.0000

R-squared 0.906515 Mean dependent var 1.60E+09
Adjusted R-squared 0.898016 S.D. dependent var 1.15E+09
S.E. of regression 3.66E+08 Akaike info criterion 42.37709
Sum squared resid 1.47E+19 Schwarz criterion 42.63126
Log likelihood -2552.814 Hannan-Quinn criter. 42.48032
F-statistic 106.6660 Durbin-Watson stat 1.998101
Prob(F-statistic) 0.000000

Inverted AR Roots .89 .00+.89i -.00-.89i -.89

Lag Distribution of ZB(-2) i Coefficient Std. Error t-Statistic

. * | 0 1043342 194609. 5.36121
. * | 1 467486. 62203.8 7.51540
. * | 2 164613. 88993.2 1.84972
. * | 3 56689.3 81182.6 0.69829
. * | 4 65684.7 70829.1 0.92737
. * | 5 113568. 93649.5 1.21269
. * | 6 122307. 98002.0 1.24800
.* | 7 13870.2 71234.7 0.19471
* . | 8 -289773. 228108. -1.27033

Sum of Lags 1757787 161965. 10.8528

However, the equation with substituted coefficients gives coefficients on the wrong lags and the coefficient on the first lag is massively too big:

Estimation Command:
=========================
LS YTXY_CORPCB C PDL(ZB(-2),8,3) @MOVAV(YTX_CORPSLSALLOC(-2),4) AR(4) @SEAS(2) @SEAS(3) @SEAS(4)

Estimation Equation:
=========================
YTXY_CORPCB = C(1) + C(2)*@MOVAV(YTX_CORPSLSALLOC(-2),4) + C(3)*@SEAS(2) + C(4)*@SEAS(3) + C(5)*@SEAS(4) + C(6)*PDL01 + C(7)*PDL02 + C(8)*PDL03 + C(9)*PDL04 + [AR(4)=C(10),ESTSMPL="1986Q1 2016Q1"]

Forecasting Equation:
=========================
YTXY_CORPCB = C(1) + C(2)*@MOVAV(YTX_CORPSLSALLOC(-2),4) + C(3)*@SEAS(2) + C(4)*@SEAS(3) + C(5)*@SEAS(4) + C(11)*ZB(-2) + C(12)*ZB(-3) + C(13)*ZB(-4) + C(14)*ZB(-5) + C(15)*ZB(-6) + C(16)*ZB(-7) + C(17)*ZB(-8) + C(18)*ZB(-9) + C(19)*ZB(-10) + [AR(4)=C(10),UNCOND,ESTSMPL="1986Q1 2016Q1"]

Substituted Coefficients:
=========================
YTXY_CORPCB = 1274786254.36 - 38878448.1674*@MOVAV(YTX_CORPSLSALLOC(-2),4) + 1705736766.64*@SEAS(2) + 748744828.564*@SEAS(3) + 402111696.221*@SEAS(4) + 1.2181288126e+17*ZB(-2) + 1043341.78828*ZB(-3) + 467486.497392*ZB(-4) + 164612.807742*ZB(-5) + 56689.3436908*ZB(-6) + 65684.7295983*ZB(-7) + 113567.589823*ZB(-8) + 122306.548724*ZB(-9) + 13870.2306607*ZB(-10) + [AR(4)=0.621460900549,UNCOND,ESTSMPL="1986Q1 2016Q1"]

Thanks.

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12292
Joined: Tue Sep 16, 2008 5:38 pm

### Re: Problem forecasting with PDL equation

Looks like a bug combining AR with PDLs. We'll fix.