Debian Codegraph @

tests/analytic_derivatives/burnside_3_order_PertParamsDerivs.mod

raw
1 % =========================================================================
2 % Copyright © 2019-2020 Dynare Team
3 %
4 % This file is part of Dynare.
5 %
6 % Dynare is free software: you can redistribute it and/or modify
7 % it under the terms of the GNU General Public License as published by
8 % the Free Software Foundation, either version 3 of the License, or
9 % (at your option) any later version.
10 %
11 % Dynare is distributed in the hope that it will be useful,
12 % but WITHOUT ANY WARRANTY; without even the implied warranty of
13 % MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 % GNU General Public License for more details.
15 %
16 % You should have received a copy of the GNU General Public License
17 % along with Dynare. If not, see <https://www.gnu.org/licenses/>.
18 % =========================================================================
19
20
21 /*
22 Check the policy functions obtained by perturbation at a high approximation
23 order, using the Burnside (1998, JEDC) model (for which the analytical form of
24 the policy function is known).
25
26 As shown by Burnside, the policy function for yₜ is:
27
28 yₜ = βⁱ exp[aᵢ+bᵢ(xₜ−xₛₛ)]
29
30
31 where:
32 θ² ⎛ 2ρ 1−ρ²ⁱ⎞
33 — aᵢ = iθxₛₛ + σ² ─────── ⎢i − ────(1−ρⁱ) + ρ² ─────⎥
34 2(1−ρ)² ⎝ 1−ρ 1−ρ² ⎠
35
36 θρ
37 — bᵢ = ───(1−ρⁱ)
38 1−ρ
39
40 — xₛₛ is the steady state of x
41 — σ is the standard deviation of e.
42
43 With some algebra, it can be shown that the derivative of yₜ at the deterministic
44 steady state is equal to:
45
46 ∂ᵐ⁺ⁿ⁺²ᵖ yₜ ∞ (2p)!
47 ──────────────── = ∑ βⁱ bᵢᵐ⁺ⁿ ρᵐ ───── cᵢᵖ exp(iθxₛₛ)
48 ∂ᵐxₜ₋₁ ∂ⁿeₜ ∂²ᵖs ⁱ⁼¹ p!
49
50 where:
51 — s is the stochastic scale factor
52
53 θ² ⎛ 2ρ 1−ρ²ⁱ⎞
54 — cᵢ = ─────── ⎢i − ────(1−ρⁱ) + ρ² ─────⎥
55 2(1−ρ)² ⎝ 1−ρ 1−ρ² ⎠
56
57 Note that derivatives with respect to an odd order for s (i.e. ∂²ᵖ⁺¹s) are always
58 equal to zero.
59
60 The policy function as returned in the oo_.dr.g_* matrices has the following properties:
61 — its elements are pre-multiplied by the Taylor coefficients;
62 — derivatives w.r.t. the stochastic scale factor have already been summed up;
63 — symmetric elements are folded (and they are not pre-multiplied by the number of repetitions).
64
65 As a consequence, the element gₘₙ corresponding to the m-th derivative w.r.t.
66 to xₜ₋₁ and the n-th derivative w.r.t. to eₜ is given by:
67
68 1 ∞ cᵢᵖ
69 gₘₙ = ────── ∑ ∑ βⁱ bᵢᵐ⁺ⁿ ρᵐ ──── exp(iθxₛₛ)
70 (m+n)! 0≤2p≤k-m-n ⁱ⁼¹ p!
71
72 where k is the order of approximation.
73
74 */
75
76 @#define ORDER = 3
77
78 var y x;
79 varobs y;
80 varexo e;
81
82 parameters beta theta rho xbar;
83 xbar = 0.0179;
84 rho = -0.139;
85 theta = -1.5;
86 theta = -10;
87 beta = 0.95;
88
89 model;
90 y = beta*exp(theta*x(+1))*(1+y(+1));
91 x = (1-rho)*xbar + rho*x(-1)+e;
92 end;
93
94 shocks;
95 var e; stderr 0.0348;
96 end;
97
98 steady_state_model;
99 x = xbar;
100 y = beta*exp(theta*xbar)/(1-beta*exp(theta*xbar));
101 end;
102
103 estimated_params;
104 stderr e, normal_pdf, 0.0348,0.01;
105 beta, normal_pdf, 0.95, 0.01;
106 theta, normal_pdf, -10, 0.01;
107 rho, normal_pdf, -0.139, 0.01;
108 xbar, normal_pdf, 0.0179, 0.01;
109 end;
110
111 steady;check;model_diagnostics;
112 stoch_simul(order=@{ORDER},k_order_solver,irf=0,drop=0,periods=0,nograph);
113 identification(order=@{ORDER},nograph,no_identification_strength);
114
115 %make sure everything is computed at prior mean
116 [xparam_prior, estim_params_]= set_prior(estim_params_,M_,options_);
117 M_ = set_all_parameters(xparam_prior,estim_params_,M_);
118 [oo_.dr,info,M_.params] = resol(0,M_, options_, oo_.dr, oo_.steady_state, oo_.exo_steady_state, oo_.exo_det_steady_state);
119
120 indpmodel = estim_params_.param_vals(:,1);
121 indpstderr = estim_params_.var_exo(:,1);
122 indpcorr = estim_params_.corrx(:,1:2);
123 totparam_nbr = length(indpmodel) + length(indpstderr) + size(indpcorr,1);
124
125 %% Verify that the policy function coefficients are correct
126 i = 1:800;
127 SE_e=sqrt(M_.Sigma_e);
128 aux1 = rho*(1-rho.^i)/(1-rho);
129 aux2 = (1-rho.^(2*i))/(1-rho^2);
130 aux3 = 1/((1-rho)^2);
131 aux4 = (i-2*aux1+rho^2*aux2);
132 aux5=aux3*aux4;
133 b = theta*aux1;
134 c = 1/2*theta^2*SE_e^2*aux5;
135
136 %derivatives wrt to rho only
137 daux1_drho = zeros(1,length(i));
138 daux2_drho = zeros(1,length(i));
139 daux3_drho = 2/((1-rho)^3);
140 daux4_drho = zeros(1,length(i));
141 daux5_drho = zeros(1,length(i));
142 for ii = 1:length(i)
143 if ii == 1
144 daux1_drho(ii) = 1;
145 daux2_drho(ii) = 0;
146 else
147 daux1_drho(ii) = rho/(rho^2 - 2*rho + 1) - 1/(rho - 1) + ((ii+1)*rho^ii)/(rho - 1) - rho^(ii+1)/(rho^2 - 2*rho + 1);
148 daux2_drho(ii) = (2*rho)/(rho^4 - 2*rho^2 + 1) + (2*ii*rho^(2*ii-1))/(rho^2 - 1) - (2*rho^(2*ii+1))/(rho^4 - 2*rho^2 + 1);
149 end
150 daux4_drho(ii) = -2*daux1_drho(ii) + 2*rho*aux2(ii) + rho^2*daux2_drho(ii);
151 daux5_drho(ii) = daux3_drho*aux4(ii) + aux3*daux4_drho(ii);
152 end
153 %derivatives of b and c wrt to all parameters
154 db = zeros(size(b,1),size(b,2),M_.exo_nbr+M_.param_nbr);
155 db(:,:,3) = aux1;%wrt theta
156 db(:,:,4) = theta*daux1_drho;%wrt rho
157 dc = zeros(size(c,1),size(c,2),M_.exo_nbr+M_.param_nbr);
158 dc(:,:,1) = theta^2*SE_e*aux3*aux4;%wrt SE_e
159 dc(:,:,3) = theta*SE_e^2*aux3*aux4;%wrt theta
160 dc(:,:,4) = 1/2*theta^2*SE_e^2*daux5_drho; %wrt rho
161
162 d2flag=0;
163 g_0 = 1/2*oo_.dr.ghs2; if ~isequal(g_0,oo_.dr.g_0); error('something wrong'); end
164 g_1 = [oo_.dr.ghx oo_.dr.ghu] +3/6*[oo_.dr.ghxss oo_.dr.ghuss]; if ~isequal(g_1,oo_.dr.g_1); error('something wrong'); end
165 g_2 = 1/2*[oo_.dr.ghxx oo_.dr.ghxu oo_.dr.ghuu]; if ~isequal(g_2,oo_.dr.g_2); error('something wrong'); end
166 g_3 = 1/6*[oo_.dr.ghxxx oo_.dr.ghxxu oo_.dr.ghxuu oo_.dr.ghuuu]; if ~isequal(g_3,oo_.dr.g_3); error('something wrong'); end
167
168 tols = [1e-4 1e-4 1e-12 1e-12];
169 KRONFLAGS = [-1 -2 0 1];
170 for k = 1:length(KRONFLAGS)
171 fprintf('KRONFLAG=%d\n',KRONFLAGS(k));
172 options_.analytic_derivation_mode = KRONFLAGS(k);
173 DERIVS = identification.get_perturbation_params_derivs(M_, options_, estim_params_, oo_.dr, oo_.steady_state, oo_.exo_steady_state, oo_.exo_det_steady_state, indpmodel, indpstderr, indpcorr, d2flag);
174 oo_.dr.dg_0 = permute(1/2*DERIVS.dghs2,[1 3 2]);
175 oo_.dr.dg_1 = cat(2,DERIVS.dghx,DERIVS.dghu) + 3/6*cat(2,DERIVS.dghxss,DERIVS.dghuss);
176 oo_.dr.dg_2 = 1/2*cat(2,DERIVS.dghxx,DERIVS.dghxu,DERIVS.dghuu);
177 oo_.dr.dg_3 = 1/6*[DERIVS.dghxxx DERIVS.dghxxu DERIVS.dghxuu DERIVS.dghuuu];
178
179 for ord = 0:@{ORDER}
180 g = oo_.dr.(['g_' num2str(ord)])(2,:); % Retrieve computed policy function for variable y
181 dg = oo_.dr.(['dg_' num2str(ord)])(2,:,:);
182 for m = 0:ord % m is the derivation order with respect to x(-1)
183 v = 0;
184 dv = zeros(1,M_.exo_nbr + M_.param_nbr);
185 for p = 0:floor((@{ORDER}-ord)/2) % 2p is the derivation order with respect to s
186 if ord+2*p > 0 % Skip the deterministic steady state constant
187 v = v + sum(beta.^i.*exp(theta*xbar*i).*b.^ord.*rho^m.*c.^p)/factorial(ord)/factorial(p);
188 %derivatives
189 dv(:,1) = dv(:,1) + sum( beta.^i.*exp(theta*xbar*i).*ord.*b.^(ord-1).*db(:,:,1).*rho^m.*c.^p...
190 +beta.^i.*exp(theta*xbar*i).*b.^ord.*rho^m.*p.*c.^(p-1).*dc(:,:,1)...
191 )/factorial(ord)/factorial(p);%wrt SE_E
192 dv(:,2) = dv(:,2) + sum( i.*beta.^(i-1).*exp(theta*xbar*i).*b.^ord.*rho^m.*c.^p...
193 +beta.^i.*exp(theta*xbar*i).*ord.*b.^(ord-1).*db(:,:,2).*rho^m.*c.^p...
194 +beta.^i.*exp(theta*xbar*i).*b.^ord.*rho^m.*p.*c.^(p-1).*dc(:,:,2)...
195 )/factorial(ord)/factorial(p);%wrt beta
196 dv(:,3) = dv(:,3) + sum( beta.^i.*exp(theta*xbar*i).*xbar.*i.*b.^ord.*rho^m.*c.^p...
197 +beta.^i.*exp(theta*xbar*i).*ord.*b.^(ord-1).*db(:,:,3).*rho^m.*c.^p...
198 +beta.^i.*exp(theta*xbar*i).*b.^ord.*rho^m.*p.*c.^(p-1).*dc(:,:,3)...
199 )/factorial(ord)/factorial(p);%wrt theta
200 dv(:,4) = dv(:,4) + sum( beta.^i.*exp(theta*xbar*i).*b.^ord.*m.*rho^(m-1).*c.^p...
201 +beta.^i.*exp(theta*xbar*i).*ord.*b.^(ord-1).*db(:,:,4).*rho^m.*c.^p...
202 +beta.^i.*exp(theta*xbar*i).*b.^ord.*rho^m.*p.*c.^(p-1).*dc(:,:,4)...
203 )/factorial(ord)/factorial(p);%wrt rho
204 dv(:,5) = dv(:,5) + sum( beta.^i.*exp(theta*xbar*i).*theta.*i.*b.^ord.*rho^m.*c.^p...
205 +beta.^i.*exp(theta*xbar*i).*ord.*b.^(ord-1).*db(:,:,5).*rho^m.*c.^p...
206 +beta.^i.*exp(theta*xbar*i).*b.^ord.*rho^m.*p.*c.^(p-1).*dc(:,:,5)...
207 )/factorial(ord)/factorial(p);%wrt xbar
208 end
209 end
210 if abs(v-g(ord+1-m)) > 1e-14
211 error(['Error in matrix oo_.dr.g_' num2str(ord)])
212 end
213 chk_dg = squeeze(dg(:,ord+1-m,:))';
214 if isempty(indpstderr)
215 chk_dv = dv(:,M_.exo_nbr+indpmodel);
216 elseif isempty(indpmodel)
217 chk_dv = dv(:,1:M_.exo_nbr);
218 else
219 chk_dv = dv;
220 end
221 fprintf('Max absolute deviation for dg_%d(2,%d,:): %e\n',ord,ord+1-m,norm( chk_dv - chk_dg, Inf));
222 if norm( chk_dv - chk_dg, Inf) > tols(k)
223 error(['Error in matrix dg_' num2str(ord)])
224 chk_dv
225 chk_dg
226 end
227 end
228 end
229 fprintf('\n');
230 end