PcGive 16: enhancements and solved problems

Fixed in PcGive 16.1

• Previous Cardt forecasts were still included after switching it off in the Forecast dialog.
• See and Also see index entries had disappeared from books.
• Better documentation of saturate PcGive batch command, and Batch in html help

New and improved in PcGive 16

• Indicators from saturation estimation (IIS etc.) are not added to the database anymore (unless they have the old name I:,S1:,T1:,DI:). Instead they are now automatically created when used; this prevents the proliferation of such variables in the database.
• Restructured saturation names and dating used in name (the date of the first deviation): DI#2020(2) II#2020(2) SI#2020(2) TI#2020(2) QI#2020(2) LS#2020(2) LT#2020(2) 2019(1) 0 0 1 -5 25 0 0 (2) 0 0 1 -4 16 0 0 (3) 0 0 1 -3 9 0 0 (4) 0 0 1 -2 4 0 0 2020(1) 0 0 1 -1 1 0 0 (2) 1 1 1 0 0 0 0 (3) -1 0 0 0 0 1 1 (4) 0 0 0 0 0 1 2 2021(1) 0 0 0 0 0 1 3 (2) 0 0 0 0 0 1 4 (3) 0 0 0 0 0 1 5 (4) 0 0 0 0 0 1 6 Note that the date indicates the first time at which the process changes.

  The set of variables has not changed, just their names, so the model estimates are unchanged:

  New name	Old name
  DI#2020(2)	DI:2020(2)
  II#2020(2)	I:2020(2)
  SI#2020(2)	S1:2020(2)
  TI#2020(2)	T1:2020(1)

• The following Algebra code can be used to create indicators in the database: algebra { "DI#2020(2)" = DI#2020(2); "II#2020(2)" = II#2020(2); "SI#2020(2)" = SI#2020(2); "TI#2020(2)" = TI#2020(2); "LS#2020(2)" = LS#2020(2); "LT#2020(2)" = LT#2020(2); "QI#2020(2)" = QI#2020(2); "MI#2020(2)@y" = MI#2020(2)@y; } algebra { "DI#2020(2)" = DI(2020,2); "II#2020(2)" = II(2020,2); "SI#2020(2)" = SI(2020,2); "TI#2020(2)" = TI(2020,2); "LS#2020(2)" = LS(2020,2); "LT#2020(2)" = LT(2020,2); "QI#2020(2)" = QI(2020,2); "MI#2020(2)@y" = MI(y,2020,2); }
• PcGive Batch has saturate("IIS"), where the argument can be: "DIS", "IIS", "SIS", "TIS", "QIS", "MIS@var", as well as the indicator name "II#", "SI#", "TI#", "I:" , "S1:", "T1:".
• More concise way to write system (old way still works), e.g.: system { Y = "log(LY)" 0:3; Z = Constant, "log(X)" 1:2 4:5 3, LC 0:5; }
• Dynamic forecasting: Automatically extending conditioning variables by Cardt forecasts (if conditional data missing in forecast period)
• Forecasting needs warning if lagged Y's are fixed: treated differently
• Write non-linear output: put names in double quotes when needed.
• Batch omitted variable test on "var" from lag 0 to lag 2: test("var", 0, 2);
• Batch has recursive() which is inserted before hedehog forecasts after recursive estimation.
• Hedgehog labelled "Dynamic forecasts (recursive estimates)" or "Dynamic simulation (full-sample estimates)" the former is used after recursive graphics.
• CATS: Ox BATCH code generated by CATS prints number of bootstrap replications.
• Forecast dialog: made number of presample observations persistent.
• Added unscaled lag weights with error bands.
• Added error bands on impulse responses (keeping inv(B) fixed).
• Arfima: printing AR and MA roots.
• Autometrics was ignoring m_dBlockSizeFraction
• Model formulation now with Back and Next buttons
• Forecasting levels: was always using crit=2

Fixed in PcGive 15

• Graph intercept adjustment after SIS on IV: Runtime error: 'matrix[1][21] XY' matrix[148][36]' bad operand
• Ox batch for static model: did not skip over missing values.
• From the PcGive test menu, it is not possible to calculate ARCH tests with lag length longer than one.
• PcGive-GARCH: Automatically generated model graph of standardized residual is using h for the standardization.
• Numerical standard errors of derived forecasts: takes the instantaneous Jacobian into account in the first SE, but not for subsequent ones.
• createinterventions batch command did not work with dated database.