Error in appliying split apply

I have a problem using splitapply for function mean.

Error using splitapply (line 111)

For N groups, every integer between 1 and N must occur at least once in the vector of group numbers.

Error in try_31032019 (line 65)

LPInew = splitapply(@mean,LPI,bins_index);

G1 = findgroups(amax2);
bins_index = discretize(G1,amaxgrid);
LPInew = splitapply(@mean,LPI,bins_index);
CSR_new = splitapply(@mean,CSR,bins_index);
constN1_60_new = splitapply(@mean,N1_60,bins_index);
constFC_new = splitapply(@mean,Const,bins_index);

Is there any one can help me to solve my problem?

[Walter Roberson]

bins_index = findgroups(discretize(G1,amaxgrid));

[Skydriver]

It has been not working right know. Is there any solution?

[Skydriver]

I have modified with this function:

Mwgrid = linspace(4.75,8.25,8);
amaxgrid = linspace(0.19,0.42,36);
G1 = findgroups(amax);
bins_index = discretize(G1,amaxgrid);
LPInew = splitapply(@mean,LPI,bins_index);
HereQL_new_1 =zeros(size(grid));
for i=1:length(LPInew)
    for j=1:length(Mwgrid)
        QL_new_1(i,j) = Coeff(1) + Coeff(2)*constN1_60 + Coeff(3).*constFC + ...
            Coeff(4).*log(LPInew(i)) + Coeff(5).*log(Mwgrid(j)) + Coeff(6).*log(constDi);
    end
end
QL_zeroes=QL_new_1; % import
QL_zeroes(QL_zeroes<0)=0; % import
PL = normcdf(QL_zeroes); 

But the bins_index become NAN for all layer.

I need to get LPInew is 36 x 1

Herewith I attach list of amax and LPI

I wish walter robertson or other can help me

[Walter Roberson]

You should not use @mean for this purpose: you should use @(x) mean(x,1) to force mean along the first dimension. Otherwise if you happen to get a group that has only a single row, then that would show up as mean(a_row_vector) which would be interpreted as mean of the vector, giving a scalar result, whereas mean() of a 2D would be mean per column giving a vector result.

[Skydriver]

And the syntax like this:

G1 = findgroups(amax);

bins_index = discretize(G1,amaxgrid);

LPInew = splitapply(@(36,mean(36,1);LPI;bins_index));?

[Akira Agata]

What you intended to do in your code is not clear for me. For example, I'm not sure how to fix the following 3 lines:

amaxgrid = linspace(0.19,0.42,36);
G1 = findgroups(amax);
bins_index = discretize(G1,amaxgrid);

Looking at your data file G1 should be integer 1~7, since amax consists of 7 different values. On the otherhand, amaxgrid becomes 1-by-36 double array between 0.19~0.42.

So I have no idea what you intended to do in the 3rd line. Could you tell us more details on what kind of calculation shall be done on your data?

[Skydriver]

I want to develop 3d figure with x absis is amaxgrid, y ordinat is Mw grid and to match with the z like state bellow. In further I have to decided that QL should be representing in the z direction with matrix of grid as 36 * 7. Yess I have small mistake to decide 7 not 8.

with this coding:

Mwgrid = linspace(4.75,8.25,7);

amaxgrid = linspace(0.19,0.42,36);

grid = meshgrid(Mwgrid,amaxgrid);

QL_new_1 =zeros(size(grid));

for i=1:length(LPInew)

for j=1:length(Mwgrid)

QL_new_1(i,j) = Coeff(1) + Coeff(2)*constN1_60 + Coeff(3).*constFC + ...

Coeff(4).*log(LPInew(i)) + Coeff(5).*log(Mwgrid(j)) + Coeff(6).*log(constDi);

end

end

QL_zeroes=QL_new_1; % import

QL_zeroes(QL_zeroes<0)=0; % import

PL = normcdf(QL_zeroes); % import

[Xq_cond1,Yq_cond1] = meshgrid(linspace(min(Mwgrid),max(Mwgrid),8),linspace(min(amaxgrid),max(amaxgrid),36)); %import

Vq_cond1 = interp2(Mwgrid, amaxgrid, PL,Xq_cond1,Yq_cond1,'cubic');

Vq_cond2 = interp2(Mwgrid, amaxgrid, PL_new,Xq_cond1,Yq_cond1,'cubic')

figure(4)

surf(Xq_cond1, Yq_cond1, Vq_cond1)

hold on

xlim([4.75 7.6])

xlabel('Magnitude (Mw)');

ylim([0 3])

ylabel('Ground Surface Acceleration (a_{max})');

zlim([0 1])

zlabel('Conditional probability of a_{max} and Mw');

Answer 1

[Mil Shastri]

The group numbers provided to splitapply need to be every integer between 1 to N. For example:

This will work:

splitapply(@mean,1:5,[2,2,3,1,1]) 

and this will not because the group number 3 is missing:

splitapply(@mean,1:5,[2,2,4,1,1])

[Christoper Angelo Malayan]

Is there a way for splitapply to SKIP this missing number and proceed to the next?

[Walter Roberson]

No.

You can use findgroups on your current non-consecutive numbers.