One approach —
V = [3.4789; 0; 0; 3.3535; 0; 0; 3.2398; 3.2263; 0; 0; 0; 3.0846; 0; 0; 2.9815; 0; 2.9430; 0; 2.9055];
r0 = [2;3;5;6;9;10;11;13;14;16;18]; %rows with 0
r1 = [1;4;7;8;12;15;17;19];
L = numel(r1);
mvc = accumarray(r1, (1:numel(r1)).', [], @(x){mean(V([r1(x) r1(min([x+1,L]))]))});
Lv = cellfun(@(x)isempty(x), mvc, 'Unif',0);
mvc([Lv{:}]) = {NaN};
mv = cell2mat(mvc);
mv = fillmissing(mv,'previous');
V([Lv{:}]) = mv([Lv{:}])
Vbfr = buffer(V,10) % Display Entire Vector In Columns
Vm = V;
V = [3.4789; 0; 0; 3.3535; 0; 0; 3.2398; 3.2263; 0; 0; 0; 3.0846; 0; 0; 2.9815; 0; 2.9430; 0; 2.9055];
Comparison = table(V, Vm, 'VariableNames',{'Original','Filled'})
First_10_Rows = Comparison(1:10,:)
Last_9_Rows = Comparison(11:end,:)
The accumarray call calculates the mean values, then computes logical vector ‘Lv’ to detect the empty cells that are then filled with NaN values. The ‘mvc’ cell array is then converted to a numerical array, and the NaN values are filled with the previous result using the fillmissing funciton. The ‘Lv’ vector us used again to fill the zeros in the original ‘V’ vector with the corresponding elements from ‘mv. to provide the result.
EDIT — (24 Feb 2024 at 15:46)
Added ‘Comparison’ table and sub-tables from it, extended the discussion of how the code works.
.
