Defined variable erroring as undefined?
显示 更早的评论
Problem:
The variable INDI is defined (checked by which search) but GJR_GARCH function still rejects it as undefined. Why?
Error:
>> GJR_GARCH
Warning: Variable names were modified to make them valid MATLAB identifiers. The original names are saved in the
VariableDescriptions property.
Undefined function or variable 'INDI'.
Error in GJR_GARCH>gloglike (line 68)
h(t) = a0+a1*eps(t-1)^2 + b1*h(t-1) + c0*INDI*eps(t-1)^2;
Error in fmincon (line 536)
initVals.f = feval(funfcn{3},X,varargin{:});
Error in GJR_GARCH (line 28)
gpar = fmincon(@gloglike,initial,sumA ,sumB,[],[],[],[],[],options,XJO_r);
Caused by:
Failure in initial objective function evaluation. FMINCON cannot continue.
Code:
function[] = garch_est()
%===================================================================
% Construct GJR_GARCH Model
%===================================================================
% read in data
formatSpec = '%s %f';
XJO = readtable('AXJO.csv','Format',formatSpec,'ReadVariableNames',true);
% calculate returns
XJO_p = XJO.AdjClose;
XJO_r = log(XJO_p(2:end)./XJO_p(1:end-1));
INDI = (XJO_r<0);
% set initial parameter estimates for optimisation
initial = [var(XJO_r)*(1-0.08-0.85-0.03);0.08;0.85;0.03];
% set constraints
sumA = [-eye(4); 0 1 1 0.5];
sumB = [-1e-9;-1e-6;-1e-6;-1e-6;0.9999];
% set options
options = optimset('Display', 'iter', 'largescale', 'off',...
'TolFun', 1e-007,'MaxIter',1000,'MaxFunEvals',1000);
% define objective function
gpar = fmincon(@gloglike,initial,sumA ,sumB,[],[],[],[],[],options,XJO_r);
% generate conditional variances
nobs = length(XJO_r);
cvar = zeros(nobs,1);
cvar(1) = gpar(1)/(1 - gpar(2) - gpar(3) - gpar(4));
for t = 2:nobs
cvar(t) = gpar(1) + gpar(2)*XJO_r(t-1)^2 + gpar(3)*cvar(t-1) + gpar(4)*INDI*XJO_r(t-1)^2;
end
pVaR = [.05];
Zscore = norminv(pVaR);
Sigma = sqrt(cvar);
Normal95 = zeros(nobs,1);
for t = 1:nobs
Normal95(t) = -Zscore*Sigma(t);
end
plot(Normal95)
end
%===================================================================
% Function for GJR_GARCH loglike
%===================================================================
function [loglike] = gloglike(initial,eps)
% estimates
a0 = initial(1);
a1 = initial(2);
b1 = initial(3);
c0 = initial(4);
n = length(eps);
h = zeros(n,1);
h(1) = a0/(1 - a1 - b1 - c0);
for t = 2:n
h(t) = a0+a1*eps(t-1)^2 + b1*h(t-1) + c0*INDI*eps(t-1)^2;
end
% Max-Loglike function and sum of
logl = - .5*log(h) - .5*eps.^2./h -.5*log(2*pi);
loglike = -sum(logl);
end
2 个评论
"Why?"
Because in MATLAB variables do not just magically jump from one workspace to another. You have to pass them explicitly:
As Walter Roberson suggested, using a nested function would be a good idea:
Geoffrey Turner
2018-10-4
采纳的回答
Walter Roberson
2018-10-4
0 个投票
INDI is defined in function garch_est, but the value for it is not passed into gloglike() . gloglike is not a nested function within garch_est so it does not share variables with garch_est .
Simplest solution: take the "end" that is after plot(Normal95) and move that "end" to after the "end" for gloglike -- so the file would have no "end" just before the function declaration for gloglike, and would have two "end" statements at the bottom of it. That would make gloglike() a nested function that would be able to share variables with garch_est()
更多回答（0 个）
类别
在 帮助中心 和 File Exchange 中查找有关 Functions 的更多信息
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
Translated by 
