Convert Matlab to Scilab

[Tranquility]

Hello every one ,

Unfortunately I'm totally new in Scilab but not in Matlab ,so I would like to convert a Matlab script to Scilab script , I tried the converter tool but it didn't work .
this is the Matlab code :

"function spectrum =spectrum-calc(data,freq,n-lags,M)
data=load ('DATA.txt');
N=size(data,1);
c-k=autocorr(data,n-lags)*var(data);%calculates the autocorrelation
c_0=var(data)%Variance
l-k=0.5*(1+cos((pi*(0:M))./M));%Tukey window calculation
for i=1:size(freq,2)
sum=0;
for e=1:M
sum=sum+l-k(1,e)*c-k(e,1)*cos(freq(1,i)*e);%Sweeps the data and generates the variance for each frequency
end
spectrum(i,1)=(c_0+2*sum)/pi ;
end
end"

Any help would be appreciated.

Tranquility

[akhorshidi]

Hi;

First off your Matlab code appears wrong because a variable name cannot contain an dash (-).

I think your function can be run in Scilab by a little modifications. So use xcorr() and xcov() instead autocorr() and cov(), respectively. See: https://help.scilab.org/docs/current/en_US/xcorr.html and https://help.scilab.org/docs/current/en_US/xcov.html

Also we can define a function in Scilab as follows:

function spectrum = spectrum-calc(data,freq,n-lags,M)

// body of the function

endfuncrion

You may need to import some data into Scilab environment, so read() function can be handy. For instance suppose we wanna load a DAT file downloaded from a website (e.g. ftp://ftp.esat.kuleuven.be/pub/SISTA/data/process_industry/powerplant.dat.gz).

So just try this:

pathname = "PATH_TO_THE_FILE";
data=read(pathname + "powerplant.dat",-1,11);
//Also You can classify your data:
u=data(:,1:5); y=data(:,6:8); yr=data(:,9:11);


HTH
Mehran
_

[Tranquility]

Hi again,
i would like to thank you for you help.
Now firstly, i wonder if you can show me how to define the n_lags of my autocorr knowing that the only information I have is that it should be < N-1 (N=size of data).Secondly, i have problem with selecting the right value of the point M which is called in signal processing the truncation point.I note that is the non parametric spectrum estimation method .
Any help would be greatly appreciated.

Tranquility

[RRogers]

N estimate: Well the old books, Nelson "applied time series analysis", give a complicated formula; but it boils down to taking the autocorrelation graph and chopping off everything beyond the point where the remainder is below .3 . This means that you graph the full n-1 lags and then use the graph to determine the significant number of lags. If you want to automate it I could give you the formula (or you could just look for .3) . If you end up with a graph that wobbles above and below the threshold you need to look at the Partial Autocorrelation graph. And so on..