MATLAB Answers

1

Delete rows when a value repeated less than 5 times

Asked by Dinh-Vinh Vo on 31 Aug 2017
Latest activity Edited by Karena Weduwen on 19 Mar 2018
I have a matrix:
a b c
1 201 1
2 202 1
3 203 1
4 204 1
5 205 1
6 206 1
7 207 2
8 208 2
9 209 2
10 210 2
11 222 3
12 232 3
I would like to remove rows 7-12 which have a value in the last column (c) repeated less than 5 times. Any help is appreciated.

  0 Comments

Sign in to comment.

2 Answers

Answer by Sean de Wolski
on 31 Aug 2017
Edited by Sean de Wolski
on 31 Aug 2017
 Accepted Answer

m = [1 201 1
2 202 1
3 203 1
4 204 1
5 205 1
6 206 1
7 207 2
8 208 2
9 209 2
10 210 2
11 222 3
12 232 3];
% Unique elements and locations in third column
[uv, ~, id] = unique(m(:,3));
% How many of each?
n = histcounts(id);
% Keep ones with more than 5.
m2 = m(ismember(m(:,3), uv(n>=5)),:)

  2 Comments

Thank you very much for your help, Sean. It is definitely what I'm looking for.
Hello Sean, your code works perfectly fine on a small testfile I use (190 rows).
But as soon as I add some more rows, the
n = histcounts(id);
part starts giving weird values. I cannot find the reason for that change.
Attached you find the two files:
working.txt code runs through without any trouble
not working.txt "n = histcounts(id);" gives strange output and the m2 stays empty.
Any suggestions on how the code could be adjusted to work on a bigger file?
since my file looks a little different (I am searching in the first column) here is the code I use:
if true
load working.txt
a = working(:,:);
[uv, ~, id] = unique(a(:,1));
n = histcounts(id);
m2 = a(ismember(a (:,1), uv(n==4)),:);
end

Sign in to comment.


Answer by José-Luis
on 31 Aug 2017
Edited by José-Luis
on 31 Aug 2017

idx = unique(m(:,3));
count = accumarray(c.',(1:numel(c)).',[],@numel);
to_keep = ~ismember(m(:,3),idx(count < 5));
result = m(to_keep,:)

  5 Comments

If I change the matrix to: m = [1 201 1 2 202 1 3 203 1 4 204 1 5 205 1 6 206 1 7 207 2 8 208 2 9 209 2 10 210 2 11 222 5 12 232 9];
It doesn't work any longer :(.
No, the solution is not robust. You'd need to adjust the accumarray function.

Sign in to comment.