Error using discretize: Bin edges must be a vector that is real, numeric or logical, and monotonically increasing.

Hi,

I have the following code:

N = numel(newdata);
split_newdata = cell(1,N);
split_newdata_mean = cell(1,N); 
for ii = 1:N
    temp = reshape([newdata{ii}{:}],[],2);
    idx = discretize(temp(:,1),[0 split_points_noID(ii,:) Inf]);
    split_newdata{ii} = groupsummary(temp(:,2),idx,@(x){x});
    split_newdata_mean{ii} = cellfun(@(x)mean(x,'omitnan'),split_newdata{ii});
end

But I continue to get the error:

Error using discretize

Bin edges must be a vector that is real, numeric or logical, and monotonically increasing.

Error in CRM_analysis (line 167)

idx = discretize(temp(:,1),[0 split_points_noID(ii,:) Inf]);

I have attached the two files here. Essentially I am trying to split the individual cells in newdata according to the cut-off points in the cows of split_points_noID. If that makes sense.

Thank you for your help!

** EDIT:

I get a really weirdly incomplete split_newdata (see file).

Answer 1

[Voss]

load newdata
load split_points_noID

Not every row of split_points_noID is monotonically increasing, because there are some trailing zeros sometimes

disp(split_points_noID)

         128         138         194         617
          76          92         155         978
          62          68         124         152
          67          71          83         155
          67          79         122         433
         199         208         269           0
          68          77          89         143
          83          88         196         322
          71          76         123         180
         135         194         256         692
         133         149         196         241
          67          71         125         186
          92         134         193         373
         121         142         242         363
          65          70          91         200
          35          62         151        1162
          73         128         198        1171
          26          34          80         335
          71          88          88           0
          79          85         142         370
          87          93         145         188
          76          80           0           0
          83         129         198         428
          86          90         133         249

To fix the code to work for this situation, remove the zeros from each row of split_points_noID, using logical indexing or nonzeros, before you use it in discretize:

N = numel(newdata);
split_newdata = cell(1,N);
split_newdata_mean = cell(1,N); 
for ii = 1:N
    temp = reshape([newdata{ii}{:}],[],2);
    idx = discretize(temp(:,1),[0 nonzeros(split_points_noID(ii,:)).' Inf]); % nonzeros
%     idx = discretize(temp(:,1),[0 split_points_noID(ii,split_points_noID(ii,:) > 0) Inf]); % logical indexing
    split_newdata{ii} = groupsummary(temp(:,2),idx,@(x){x});
    split_newdata_mean{ii} = cellfun(@(x)mean(x,'omitnan'),split_newdata{ii});
end

[lil brain]

Hi @Voss I seem to still get the same error.

Error using discretize

Bin edges must be a vector that is real, numeric or logical, and monotonically increasing.

Error in CRM_analysis (line 170)

idx = discretize(temp(:,1),[0 nonzeros(split_points_noID(ii,:)).' Inf]); % nonzeros

I tried both nonzeros and logical indexing.

[Voss]

Maybe you are using a different split_points_noID because they both run here:

load newdata
load split_points_noID

N = numel(newdata);
split_newdata = cell(1,N);
split_newdata_mean = cell(1,N); 
for ii = 1:N
    temp = reshape([newdata{ii}{:}],[],2);
    idx = discretize(temp(:,1),[0 split_points_noID(ii,split_points_noID(ii,:) > 0) Inf]); % logical indexing
    split_newdata{ii} = groupsummary(temp(:,2),idx,@(x){x});
    split_newdata_mean{ii} = cellfun(@(x)mean(x,'omitnan'),split_newdata{ii});
end
split_newdata

split_newdata = 1×24 cell array
    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {4×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {3×1 cell}    {5×1 cell}    {5×1 cell}    {3×1 cell}    {5×1 cell}    {5×1 cell}

N = numel(newdata);
split_newdata = cell(1,N);
split_newdata_mean = cell(1,N); 
for ii = 1:N
    temp = reshape([newdata{ii}{:}],[],2);
    idx = discretize(temp(:,1),[0 nonzeros(split_points_noID(ii,:)).' Inf]); % nonzeros
    split_newdata{ii} = groupsummary(temp(:,2),idx,@(x){x});
    split_newdata_mean{ii} = cellfun(@(x)mean(x,'omitnan'),split_newdata{ii});
end
split_newdata

split_newdata = 1×24 cell array
    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {4×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {5×1 cell}    {3×1 cell}    {5×1 cell}    {5×1 cell}    {3×1 cell}    {5×1 cell}    {5×1 cell}

[lil brain]

You're right it does work. Something was off with my version. Thanks!

[Steven Lord]

If you encounter this problem again I would set an error breakpoint and run your code. When MATLAB stops at the breakpoint, ask:

issorted(split_points_noID(ii, :))

If that returns false, you need to investigate what's on row ii of split_points_noID.