MATLAB: Find maximum values related to index

findfor loopindexMATLABmax

Hello community,

I am trying to write a for loop to find the maximum values in column 2 for each group variable in column 1. I tried a for loop to no avail. Any help would be appreciated.

The solution should result in a maximum value, corresponding row index, for each group variable in column 1. So, the solution should yield 1) the group index associated with stests(:,1) 2) maximum values of stests(:,2) per group index 3) the row index of stests corresponding to each max value result

I attached a 2418×2 matrix called stests.

I apologize ahead of time for my lack of proper data science language.

Very much appreciated.

Regards, Mark

Best Answer

Ideal job for findgroups and splitapply.

VAL is the max value in each group, located at id_l (local group index). For example, id_l=2 if max value is second value in group.

data=xlsread('stests.xlsx');
%%Find max value and local index
G=findgroups(data(:,1));
[VAL,id_l]=splitapply(@max,data(:,2),G);
%%Find global index
[~,b]=unique(data(:,1));
id_g=[id_l+b]-1;
%Store results
out=[VAL,id_g]

Related Solutions

MATLAB: How do you get multiple minimum values in a column based on grouping and then isolate the corresponding rows for the whole table

[g,ig]=findgroups(data.names);
b=splitapply(@(x) {maxk(x,10)},data.age,g);

b contains the k oldest elements for each group...you'll have to use an m-file function to return the optional second index from maxk to locate the position of the N individuals in the group.

>> [cellstr(ig) b]
ans =
  7×2 cell array
    {'Ben'  }    {10×1 double}
    {'Billy'}    { 4×1 double}
    {'Happy'}    { 4×1 double}
    {'Jenny'}    {10×1 double}
    {'Joe'  }    { 8×1 double}
    {'Kate' }    {10×1 double}
    {'Smith'}    {10×1 double}
>>

ADDENDUM: The indexing would look something like--

function [mx,ix]=getmaxk(x,n)
% wrapper for maxk cuz can't return second output w/ anonymous function
  [mx,ix]=maxk(x,n);
  mx={mx};
  ix={ix};
end
fnmaxk=@(x) getmaxk(x,10);
[b,ib]=splitapply(fnmaxk,data.age,g);

ADDENDUM SECOND:

OK, it's not so bad after all -- have to do it explicitly right now; seems like a good candidate for an enhancement to splitapply() to have available as an auxiliary variable.

Start with

function sq=getmaxkgrp(x,v)
% returns the group variable sequence associated with the n maxk values of x
  n=10;
  [mx,ix]=maxk(x,n);
  sq=v(ix);
end

which has the 10 value hardcoded; you can use the indirect method illustrated above to be able to change the number for which you're looking.

Then, as suggested, augment the table with

data.seq=[1:height(data)].';        % add the sequence number in table of each row
ix=splitapply(@(x1,x2) {getmaxkgrp(x1,x2)},data.age,data.seq,g);  % and return that seq number for the top n
tmaxk=(cellfun(@(ix) data(ix,1:end-1),ix,'uni',0));  % retrieve those records from full table
tmaxk=vertcat(tmaxk{:});                             % and combine into a table

Above results in--

>> tmaxk =
  56×3 table
    names    age    pet 
    _____    ___    ____
    Ben      31     fish
    Ben      31     bird
    Ben      31     bird
    Ben      31     bird
    Ben      31     fish
    Ben      21     dog 
    Ben      21     fish
    Ben      21     dog 
    Ben      21     fish
    Ben      17     dog 
    Billy    30     fish
    Billy    30     bird
    Billy    30     fish
    Billy    30     bird
    Happy    31     bird
    Happy    31     cat 
    Happy    31     bird
    Happy    31     cat 
    Jenny    30     cat 
    Jenny    30     dog 
    Jenny    30     dog 
    Jenny    30     fish
    Jenny    30     dog 
    Jenny    30     fish
    Jenny    30     dog 
    Jenny    30     fish
    Jenny    30     cat 
    Jenny    30     dog 
    Joe      79     dog 
    Joe      79     dog 
    Joe      21     dog 
    Joe      20     cat 
    Joe      19     cat 
    Joe      18     bird
    Joe      15     cat 
    Joe      15     cat 
    Kate     60     fish
    Kate     60     fish
    Kate     60     fish
    Kate     51     fish
    Kate     51     fish
    Kate     51     fish
    Kate     51     fish
    Kate     50     bird
    Kate     50     dog 
    Kate     50     bird
    Smith    38     fish
    Smith    38     bird
    Smith    38     fish
    Smith    38     bird
    Smith    38     bird
    Smith    38     fish
    Smith    30     bird
    Smith    29     bird
    Smith    29     cat 
    Smith    29     cat 
>>

I didn't test it extensively but I believe it will have returned the correct values. Using a smaller value for N would make that part much easier; left as "Exercise for Student!" :)

MATLAB: Index of maximum values in accumarray

A better approach might be to use the splitapply to divide the data based on days. Then the better will be better organized and you can easily process for required results. For example

splitData = splitapply(@(x) {x}, wind(:,2:3), findgroups(c));

now you can search splitData to find the max value and the corresponding direction,

result = cellfun(@(x) x(x(:,1)==max(x(:,1)), :), splitData, , 'UniformOutput', 0);

the result will contain the maximum daily value and wind direction.

Best Answer

Related Solutions

MATLAB: How do you get multiple minimum values in a column based on grouping and then isolate the corresponding rows for the whole table

MATLAB: Index of maximum values in accumarray

Related Question