I have 4 different dimension matrices(A:1100X1, B:2200X1,C: 3300X1,D:3300X1)..I want to keep all 4 in one matrix..

Hi, I have 4 different dimension matrices(A:1100X1, B:2200X1,C: 3300X1,D:3300X1)..I want to keep all 4 in one matrix..

the code written as

clc 
clear all 
format compact 
warning off 
L = []; indexgood=1; 

%mrsgarch_n

load mrsgarch_n 

eval(['L(:,',num2str(indexgood),') = Loglike(:);']); 

indexgood=indexgood+1;

% MRSGARCH-t2 

load mrsgarch_t2 

eval(['L(:,',num2str(indexgood),') = Loglike(:);']); 

indexgood=indexgood+1;

% MRSGARCH-GED 

load mrsgarch_g 

eval(['L(:,',num2str(indexgood),') = Loglike(:);']); 

indexgood=indexgood+1;

% MRSGARCH-t1 

load mrsgarch_t 

eval(['L(:,',num2str(indexgood),') = Loglike(:);']); 

indexgood=indexgood+1;

cd junk\VF

save loglik.txt L -ascii -double -tabs

cd ..

cd ..

when I run this code.. the following error msg is coming..

??? Subscripted assignment dimension mismatch.

Error in ==> ooslikelihoods at 18 eval(['L(:,',num2str(indexgood),') = Loglike(:);']);

Thanks

Kishore

[Fangjun Jiang]

What are you trying to do? Your code looks awkward. How does your code relate to the question in your title?

[Kishore Dhavala]

Thanks Jiang for the comment..

I am new to Matlab, I have downloaded a program prepared by Juri Marcucci and trying to replicate the tables presented in his paper "Forecasting Stock Market Volatility with Regime-Switching GARCH Models".

When run the program for insample, it stored likelihood ratios of 4 different distributions in the form of mrcgarch_n.mat, mrsgarch_t.mat, mrsgarch_t2.mat and mrsgarch_g.mat, all have different rowsX1 matrices.

the current file to is get the loglikelihoods for the out of sample evaluations for that I need to get all the log-likelihoods into one matrix.

The above program is written by him, the program is running and storing the all the likelihoods when I give different names to L..

load mrsgarch_n

eval(['L(:,',num2str(indexgood),') = Loglike(:);']);

indexgood=indexgood+1;

% MRSGARCH-t2

load mrsgarch_t2

eval(['L1(:,',num2str(indexgood),') = Loglike(:);']);

indexgood=indexgood+1;

% MRSGARCH-GED

load mrsgarch_g

eval(['L2(:,',num2str(indexgood),') = Loglike(:);']);

indexgood=indexgood+1;

% MRSGARCH-t1

load mrsgarch_t

eval(['L3(:,',num2str(indexgood),') = Loglike(:);']);

indexgood=indexgood+1;

Thanks

[Walter Roberson]

That's either a good example of bad programming, or else code written for MATLAB 4 or earlier.

Answer 1

[Fangjun Jiang]

I assume each of your data file contains the same variable Loglike, but the size is different from file to file, first is 1100x1, second is 2200x1, etc. You want to combine them.

Try this:

L = []; 
load mrsgarch_n
L=[L;Loglike]; 
load mrsgarch_t2
L=[L;Loglike];  
load mrsgarch_g
L=[L;Loglike];
load mrsgarch_t
L=[L;Loglike];

[Kishore Dhavala]

Thanks Jiang, will try this.

Answer 2

[Walter Roberson]

First, please read:

http://matlab.wikia.com/wiki/FAQ#How_can_I_create_variables_A1.2C_A2.2C....2CA10_in_a_loop.3F

http://matlab.wikia.com/wiki/FAQ#Why_is_it_advised_to_avoid_using_the_.22eval.22_function.3F

Beyond that: it is not possible in MATLAB to have numeric arrays that have different column sizes. If you want to use save() specifically to create four columns of data of different sizes that load() would load in, then you cannot do that. There are other ways that it would be possible to get the data written out in to columns, but load() would not be able to handle the result.

Does your data need to be saved as text? If so, then what are the requirements imposed by what-ever program you plan to read it in with? It makes more sense to construct a solution that fits the needs of what you cannot change, then for us to suggest different things you might want to do.

[Kishore Dhavala]

Thanks Walter..Yes I want to save these results in a textfile and call it in another program.

[Walter Roberson]

Do you have control of how the other program reads the data? Or does it need a particular format? If it needs a particular format, what is that format? If it is expecting spaces to indicate that there are no more values in a column, then it is probably expecting each column to be a very specific width.

Is the other program able to treat NaN as indicating that there is no data?