problem using polyfitn with for loop and nchoosek

hi, i have this problem, i need to create with my matrix a for loop with polyfitn using nchoosek with every 2 single combination of column vectors.

my matrix is ( every column[loudness,flustr.....] is the variable x of polyfitn):

Loudness=[2.79;3.16;3.71;2.29;2.49;2.64;2.9;2.79;2.91;3.35];

FlucStr=[0.0256;0.0277;0.0311;0.0246;0.021;0.0199;0.0194;0.0256;0.0213;0.0208];

Roughness=[0.491;0.6;0.728;0.34;0.425;0.515;0.617;0.491;0.389;0.438];

Sharpness=[1.03;1.11;1.21;0.887;0.934;0.954;0.985;1.03;1.04;1.12];

Leq=[39.7;40.9;42.6;38.1;38.9;39.5;40.6;39.7;40.3;41.7];

SIL=[29.4;30.9;32.9;26.9;28;28.8;30.1;29.4;28.8;30];

Tonality=[0.133;0.128;0.113;0.153;0.14;0.131;0.118;0.133;0.203;0.18];

Kurtosis=[2.2;2.2;2.2;2.44;2.49;2.48;2.45;2.2;2.39;2.38];

metriche=[Loudness FlucStr Roughness Sharpness Leq SIL Tonality Kurtosis];

and subjective is(is the variable y of polyfitn):

subjective=[7.5;7.02;6.94;7.91;7.96;7.91;7.78;7.42;7.86;7.47];

how can i calculate polyfitn using a for loop and nchoosek???

Answer 1

[Rishabh Mishra]

Hi mohamed,

I assume that

• You are using the matrix ‘metriche’ as input data for the ‘polyfitn’ function.
• The vector ‘subjective’ is used as output data for ‘polyfitn’ function.
• You want to apply ‘polyfitn’ function on each combination of 2 columns out of the 8 columns of ‘metriche’ using for loop, ‘nchoosek’ & ‘polyfitn’ functions.

Refer to the code below:

% create the required matrix from given columns
metriche = [Loudness FlucStr Roughness Sharpness Leq SIL Tonality Kurtosis]
% 'sz' stores number of columns in 'metriche'
number_of_columns = size(metriche,2)
% 'indices' stores an array of column numbers 
indices = 1:number_of_columns
 
% 'idx' is a vector
% 'idx' stores all the column combinations when 2 columns are chosen at a
% time out of given 8 columns
idx = nchoosek(indices,2)
 
% use for loop to evaluate polyfit for each of the column combinations
for k = 1:size(idx,1)
    % the polyfit is linear in nature (1-degree)
    degree = 1
    % apply polyfitn on combinations of 2 columns from 'metriche' as inputs
    % 'subjective' is the output vector
    % the polynomial fir is a 1-Degree fit
    evaluated_polyfit = polyfitn(metriche(:,[idx(k,1) idx(k,2)]), subjective, degree)
end

For better understanding, refer to documentation of ‘nchoosek’ function in this link, and refer to the section on how combinations of elements are generated from vectors.

Hope this helps.