Smartly search between tables

Hi all,

so I have two datasets made of 0 and 1. The two datasets (in the form of matlab tables), call them D1 and D2 are 1229x119. D1 contains true values and D2 predictions. Each column of D1 and D2 has a country name c1,c2,c3,...,c119. What I would like to do is to recover, for each country in a subset, say [c3,c6,c12,c45,c117], the row index for these countries whenever D2 has entry 1 and D1 has entry 0.

The desired output is a cell array indexed by country (say row_index{ck} for k=3,6,12,45,117) whose values are the row indices for which D2=1 and D1=0, s for instance say this happens fr rows 39, 678, 1220 for cuntry c6, I would like too obtain:

row_index{c6}=[39,678,1220]

Could you please help me out on this?

[dpb]

That's a straightforward logical addressing expression, but I'd suggest you may not need to build the explicit index but use rowfun with the Country as grouping variable and the anonymous (or m-file) function operating on the selected subset doing the conditional test on the D variables -- or just build a D1D2 variable that is the logical AND of the two which would be (D1==0).*(D2==1)

[federico nutarelli]

@dpb thank you for the reply. Actually I am new to MATLAB. I am using the follwing code:

[row_indices(k),col_indices(k),vals(k)] = find(mcv(:,rich_countries(k))=1 & ttv(:,rich_countries(k))=0);

but it doess not seem to work. Here D1 is mcv and D2 is tttv. rich_countries(k) represents the subset of countries

[Peter Perkins]

dpb, it sounds like the countries are the variables in these tables, so I would think this is a varfun problem, not rowfun.

Answer 1

[Peter Perkins]

This would be trivial if varfun accepted more than one table input, but it doesn't do that. But you can make one table that looks for your condition, and apply varfun to that:

>> D1 = array2table(randi([0 1],10,3),"VariableNames",["US" "UK" "UAE"]);
>> D2 = array2table(randi([0 1],10,3),"VariableNames",["US" "UK" "UAE"]);
>> D12 = array2table(D1{:,:} & ~D2{:,:})
D12 =
  10×3 table
    Var1     Var2     Var3 
    _____    _____    _____
    true     true     false
    false    false    false
    true     false    false
    false    false    false
    false    false    false
    false    false    true 
    false    true     true 
    true     true     true 
    true     true     true 
    false    true     false
>> varfun(@(x) find(x), D12, "OutputFormat","cell")
ans =
  1×3 cell array
    {4×1 double}    {5×1 double}    {4×1 double}

1229x119 isn't gigantic, so probably no memory issues there. But alternatively,

for i = 1:width(D1)
    c{i} = find(D1.(i) & ~D2.(i));
end