Arrary indices must be positive integers or logical values

for t=TestWindow
i=t-TestWindowStart+1;
EstimationWindow=t-EstimationWindowSize:t-1;
Sigma=std(Returns(EstimationWindow));
Normal95(i)=-Zscore(1)*Sigma;
Normal99(i)=-zscore(2)*Sigma;
end

Array indices must be positive integers or logical values.

Answer 1

[Cris LaPierre]

One of your indices is <= 0. You haven't shared enough information for us to say exactly what the problem is.

See Ch 5 of MATLAB Onrmap to learn how to properly index an array.

[Thukten Tashi]

I am trying to calculate value at risk. I have followed the steps as provied in https://uk.mathworks.com/help/risk/value-at-risk-estimation-and-backtesting.html, however it isn't working on my data. Please help. Here is the syntax that I used:

>> Returns=tick2ret(Price);
>> DateReturns=Date(2:end,:);
>> SampleSize=length(Returns);
>> 
>> %%Deteremining 250 days
>> TestWindowStart=find(year(DateReturns)==2016,1);
>> TestWindow=TestWindowStart:SampleSize;
>> EstimationWindowSize=250;
>> 
>> %%Confidence level
>> pVaR=[0.05 0.01];
>> 
>> %%Normal method
>> Zscore=norminv(pVaR);
>> Normal95=zeros(length(TestWindow),1);
>> Normal99=zeros(length(TestWindow),1);
>> 
>> for t=TestWindow
i=t-TestWindowStart+1;
EstimationWindow=t-EstimationWindowSize:t-1;
Sigma=std(Returns(EstimationWindow));
Normal95(i)=-Zscore(1)*Sigma;
Normal99(i)=-Zscore(2)*Sigma;
end

Array indices must be positive integers or logical values.

[Thukten Tashi]

[Cris LaPierre]

It's an indexing error, so check the value of the variables you are using for indexing. At least one of them is <=0, which is not allowed.

[Walter Roberson]

EstimationWindow=t-EstimationWindowSize:t-1;

your t is 1. EstimationWindowSize is 250. 1-250:1-1 is -249 to 0.

[Walter Roberson]

That is, the estimation window is looking before the current point, and so that fails if the current point is not at least EstimationWindowSize into the data.

[Thukten Tashi]

How should I put it please? I am trying to estimate VaR on 251st day using the past 250 days and then continue the same for all the remaing days. I tried several ways, it doesn't seem to work. Your help would be highly appreciated.

[Cris LaPierre]

Then start t at 251, not 1.

[Thukten Tashi]

Still getting wrong. Although I am using 250 days to calculate one day Var on 251st day, the acutal days in some of the years is less than 250. I don't know how to account of that. Using t=251, I am still getting error. Here is the syntax.

>> Returns=tick2ret(Price);
DateReturns=Date(2:end,:);
SampleSize=length(Returns);
>> 
>> 
>> TestWindowStart=find(year(DateReturns)==2017,251);
>> TestWindow=TestWindowStart:SampleSize;
EstimationWindowSize=250;
>> 
>> 
>> pVaR=[0.05 0.01];
>> Zscore=norminv(pVaR);
Normal95=zeros(length(TestWindow),1);
>> Normal95=zeros(length(TestWindow),251);
>> Normal99=zeros(length(TestWindow),251);
>> 
>> for t=TestWindow
i=t-TestWindowStart+1;
EstimationWindow=t-EstimationWindowSize:t-1;
Sigma=std(Returns(EstimationWindow));
Normal95(i)=-Zscore(1)*Sigma;
Normal99(i)=-Zscore(2)*Sigma;
end
Array indices must be positive integers or logical values.

[Walter Roberson]

For one thing, do not take length() when you do not know for sure which dimension will be shorter. Use size() of the dimension you are interested in.

length(X) is defined as:

temporary = size(X)
if any(temporary==0)
   length is 0
else
    length is max(temporary)
end

[Thukten Tashi]

I tried this way. I don't know if the VaR has been calculated as I couldn't plot graph.

[Walter Roberson]

Notice your TestWindowStart is a vector of length 230 -- you have 230 entries with year 2017. Your find() asked for a maximum of 251 such entries, which is suspicious .

Notice that you did not ask for the first such entry: you asked for at most 251 of them, no matter how they are scattered through the data. You did not make any attempt to group them either, for the possibility that you had (say) 100 here and another 130 over there, that you wanted to (hypothetically) treat as two groups.

I suspect that your dates are sorted, and that you want the first such entry rather than up to 251 such entries.

[Thukten Tashi]

Thank you so much for your help. I managed to calculate somehow.

[Thukten Tashi]

I have calculated VaR using exponential weight moving average. The figure for first failire under EWMA at 95% confidence level appears as '1'. I thougth that this value should be 77 instead. I am not sure which one is correct. I have used the following steps.

Lambda = 0.94;

Sigma2 = zeros(length(Returns),1);

Sigma2(1) = Returns(1)^2;

for i = 2 : (TestWindowStart-1)

Sigma2(i) = (1-Lambda) * Returns(i-1)^2 + Lambda * Sigma2(i-1);

end

Zscore = norminv(pVaR);

EWMA95 = zeros(length(TestWindow),1);

EWMA99 = zeros(length(TestWindow),1);

for t = TestWindow

k = t - TestWindowStart + 1;

Sigma2(t) = (1-Lambda) * Returns(t-1)^2 + Lambda * Sigma2(t-1);

Sigma = sqrt(Sigma2(t));

EWMA95(k) = -Zscore(1)*Sigma;

EWMA99(k) = -Zscore(2)*Sigma;

end