Method to count number of observations by category

What would be the best method to count observations which fit the criteria number listed in the last column:

>> s

s =

     22.9924   46.8888    1.4597    4.8547    1.3379    3.8795   24.8686    3.0000
     19.7965   40.1072    1.4644    5.2458    1.3262    3.0253   19.7573    3.0000
     12.0315   70.3394    1.8392    3.6708    1.2664    2.2959   13.2066    3.0000
      0.0854    4.4000    1.3399    1.6076    1.2214    1.9499    1.7144    2.0000
    -10.7796   26.7116    1.0871    1.0044    1.2722    1.6683  -12.4818    1.0000
      1.3617   15.7695    1.0896    1.4550    1.2444    1.9098    1.2284    2.0000
     14.2101   11.4473    1.2464    4.0444    1.2413    2.6900   16.6996    3.0000
     12.9187   60.9303    1.5128    3.4454    1.2593    2.2976   15.4214    3.0000
      4.8807   11.4097    1.4971    2.0412    1.2105    1.7832    5.7704    2.0000
      5.7536   22.7122    1.1432    1.8351    1.2305    2.1097    9.1543    3.0000

another example:

s =

     -3.8660    3.8000    1.8212    1.0659    0.9359    1.1195   -4.2488    4.0000
      5.6911    1.6000    1.8920    1.0000    1.0772    1.1631    7.5085    6.0000
      1.6582   31.9250    1.1889    1.0221    1.1091    1.1899    2.0074    5.0000
     -3.0589    1.1000    1.6842    1.0184    0.9804    1.1165   -4.5364    4.0000

Such that I will have an alert that I only have one observation of , say 1.00 in the first piece of data, but 5 and 6 in the second piece?

Thanks, Kate

Answer 1

[Cedric]

How about

 counts = accumarray( s(:,end), 1 ) ;

In the first case, it gives

 counts =
     1
     3
     6

In the second case

 counts =
     0
     0
     0
     2
     1
     1

As you can see, the i-th element of vector counts is the count of data i. And then you can process counts as you want/need, e.g.

 is1 = counts == 1 ;
 if any( is1 )
     fprintf( 'Warning, the following data blabla..\n' ) ;
     disp( find( is1 )) ;
 end

[Kate]

Thanks Cedric, another great use of accumarray.

Answer 2

[Image Analyst]

I believe this is what you want:

s =[...
     22.9924   46.8888    1.4597    4.8547    1.3379    3.8795   24.8686    3.0000
     19.7965   40.1072    1.4644    5.2458    1.3262    3.0253   19.7573    3.0000
     12.0315   70.3394    1.8392    3.6708    1.2664    2.2959   13.2066    3.0000
      0.0854    4.4000    1.3399    1.6076    1.2214    1.9499    1.7144    2.0000
    -10.7796   26.7116    1.0871    1.0044    1.2722    1.6683  -12.4818    1.0000
      1.3617   15.7695    1.0896    1.4550    1.2444    1.9098    1.2284    2.0000
     14.2101   11.4473    1.2464    4.0444    1.2413    2.6900   16.6996    3.0000
     12.9187   60.9303    1.5128    3.4454    1.2593    2.2976   15.4214    18.0000
      4.8807   11.4097    1.4971    2.0412    1.2105    1.7832    5.7704    2.0000
      5.7536   22.7122    1.1432    1.8351    1.2305    2.1097    9.1543    3.0000]

category = s(:, end); % Extract last column
% Get histogram of it.
edges = unique(category); % Edges of histogram
counts = histc(category, unique(category))
% Find indexes for which the count is 1.
singleCounts = find(counts == 1)
for k = 1 : length(singleCounts)
  thisIndex = singleCounts(k);
  numberAtIndex = edges(thisIndex);
  message = sprintf('The category %d shows up only once!', numberAtIndex);
  uiwait(helpdlg(message));
end

It takes the histogram of the last column. Any number that shows up in that column only one time will popup a message to the user. I changed your s slightly so that the number 1 and the number 18 each show up only once in the last column.

[Kate]

Thanks, but I'm actually looking for ANY value which is only seen one time, not just observations of s==1. See above discussion.

[Image Analyst]

Did you actually run my code? Maybe you ran it before I added the edit to have 18 also show up only once. Run it all the way through. You will find that it pops up two messages. The first one says that 1 shows up only once, and the second one says that 18 shows up only once. Isn't that what you got?

[Image Analyst]

Kate, my code runs fine on your new, additional example with 5 and 6. Here, look at this:

s =[...
  -3.8660    3.8000    1.8212    1.0659    0.9359    1.1195   -4.2488    4.0000
  5.6911    1.6000    1.8920    1.0000    1.0772    1.1631    7.5085    6.0000
  1.6582   31.9250    1.1889    1.0221    1.1091    1.1899    2.0074    5.0000
  -3.0589    1.1000    1.6842    1.0184    0.9804    1.1165   -4.5364    4.0000]
category = s(:, end); % Extract last column
% Get histogram of it.
edges = unique(category); % Edges of histogram
counts = histc(category, edges )
% Find indexes for which the count is 1
singleCounts = find(counts == 1)
for k = 1 : length(singleCounts)
  thisIndex = singleCounts(k);
  numberAtIndex = edges(thisIndex);
  message = sprintf('The category %d shows up only once!', numberAtIndex);
  fprintf('%s\n', message);
  uiwait(helpdlg(message));
end

In addition to the popup message box, you'll see this in the command window:

s =
   -3.8660    3.8000    1.8212    1.0659    0.9359    1.1195   -4.2488    4.0000
    5.6911    1.6000    1.8920    1.0000    1.0772    1.1631    7.5085    6.0000
    1.6582   31.9250    1.1889    1.0221    1.1091    1.1899    2.0074    5.0000
   -3.0589    1.1000    1.6842    1.0184    0.9804    1.1165   -4.5364    4.0000
counts =
     2
     1
     1
singleCounts =
     2
     3
The category 5 shows up only once!
The category 6 shows up only once!

Tell me if this doesn't do exactly and perfectly what you want to do and you asked for.

[Kate]

Thanks Image Analyst, the new code works. I had run it on your previous code when I replied.

Answer 3

[Ben11]

what about:

size(s(s == 1.00),1))

which would give you the number of occurences in s meeting the criterion.

Or maybe something like this

if any(s(s == 1.00))

msgbox('we have a winner')

end

[Kate]

Is there a method where you can avoid specifying what s equals(e.g. s(s==1.00)? In some instances I might have only one observation of 3 or 18, etc etc.

[Ben11]

Ok I'm not quite sure to understand sorry. If we take your matrix above for example, would you like to know how many times the value in the last column is present in the corresponding row? Or in the whole matrix? Or would you like to know that there are six 3's, three 2's and one 1?

[Kate]

To illustrate my question, here is a different piece of data. Here I would want to know that 5 and 6 are only seen once:

s =

     -3.8660    3.8000    1.8212    1.0659    0.9359    1.1195   -4.2488    4.0000
      5.6911    1.6000    1.8920    1.0000    1.0772    1.1631    7.5085    6.0000
      1.6582   31.9250    1.1889    1.0221    1.1091    1.1899    2.0074    5.0000
     -3.0589    1.1000    1.6842    1.0184    0.9804    1.1165   -4.5364    4.0000