Control Break Processing

[EDIT: 20110720 16:05 CDT - reformat - WDR]

I have 3 column vectors (Date1 Date2 and Error) and what I want to do is create the column named Target. Target is a 1 if it is the smallest error by Date1 and Date2 AND error is < .10 otherwise its a 0.

I can do this using for loops, BUT I was wondering if I can do this using logical indexing....I am guessing that this will be faster since I won't have to nest3 for loops.

Date1             Date2  error  Target
'06-Nov-2000'  '16-Dec-2000'  0.47  0
'06-Nov-2000'  '16-Dec-2000'  0.41  0
'06-Nov-2000'  '16-Dec-2000'  0.29  0
'06-Nov-2000'  '16-Dec-2000'  0.14  0
'06-Nov-2000'  '16-Dec-2000'  0.03  1
'06-Nov-2000'  '16-Dec-2000'  0.18  0
'06-Nov-2000'  '18-Nov-2000'  0.03  1
'06-Nov-2000'  '18-Nov-2000'  0.31  0
'06-Nov-2000'  '18-Nov-2000'  0.44  0
'06-Nov-2000'  '18-Nov-2000'  0.45  0
'06-Nov-2000'  '18-Nov-2000'  0.44  0
'06-Nov-2000'  '18-Nov-2000'  0.44  0
'06-Nov-2000'  '18-Nov-2000'  0.45  0
'06-Nov-2000'  '18-Nov-2000'  0.45  0
'07-Nov-2000'  '18-Nov-2000'  0.36  0
'07-Nov-2000'  '18-Nov-2000'  0.09  1
'07-Nov-2000'  '18-Nov-2000'  0.15  0
'07-Nov-2000'  '18-Nov-2000'  0.38  0
'07-Nov-2000'  '18-Nov-2000'  0.48  0
'07-Nov-2000'  18-Dec-2000'  0.50  0
'07-Nov-2000'  18-Dec-2000'  0.01  1
'07-Nov-2000'  18-Dec-2000'  0.15  0

Answer 1

[Fangjun Jiang]

I have one solution but it still requires a for-loop.

raw={'06-Nov-2000'  '16-Dec-2000'  0.47  0
'06-Nov-2000'  '16-Dec-2000'  0.41  0
'06-Nov-2000'  '16-Dec-2000'  0.29  0
'06-Nov-2000'  '16-Dec-2000'  0.14  0
'06-Nov-2000'  '16-Dec-2000'  0.03  1
'06-Nov-2000'  '16-Dec-2000'  0.18  0
'06-Nov-2000'  '18-Nov-2000'  0.03  1
'06-Nov-2000'  '18-Nov-2000'  0.31  0
'06-Nov-2000'  '18-Nov-2000'  0.44  0
'06-Nov-2000'  '18-Nov-2000'  0.45  0
'06-Nov-2000'  '18-Nov-2000'  0.44  0
'06-Nov-2000'  '18-Nov-2000'  0.44  0
'06-Nov-2000'  '18-Nov-2000'  0.45  0
'06-Nov-2000'  '18-Nov-2000'  0.45  0
'07-Nov-2000'  '18-Nov-2000'  0.36  0
'07-Nov-2000'  '18-Nov-2000'  0.09  1
'07-Nov-2000'  '18-Nov-2000'  0.15  0
'07-Nov-2000'  '18-Nov-2000'  0.38  0
'07-Nov-2000'  '18-Nov-2000'  0.48  0
'07-Nov-2000'  '18-Dec-2000'  0.50  0
'07-Nov-2000'  '18-Dec-2000'  0.01  1
'07-Nov-2000'  '18-Dec-2000'  0.15  0};

% Date1 is your first column, nx1 cell array of strings
% Date2 is your second column, nx1 cell array of strings.
% Err is your third column, nx1 double array

Date1=raw(:,1);
Date2=raw(:,2);
Err=cell2mat(raw(:,3));
RawTarget=cell2mat(raw(:,4));

Target=false(size(Err));
Date=cellstr(cell2mat([Date1,Date2]));
UniqueDate=unique(Date);
for k=1:length(UniqueDate)
    Index=ismember(Date,UniqueDate(k));
    Loc=find(Index);
    [Dummy,MinIndex]=min(Err(Index));
    Target(Loc(MinIndex))=true;
end
Target=and(Target, Err<0.1);
all(Target==RawTarget)

[Nathan Greco]

Note that you should avoid using "error" as a variable name, since it is a Matlab function. Use err or something similar to get your point across in the future.

[Fangjun Jiang]

Good point, thank you! I'll make the change.

[luke]

I am not sure that this is doing what it is suppose to. If you look at the sample file above, you will notice that there is a Target 1 for Date1 06-Nov-2000 Date2 16-Dec-2000 AND Date1 06-Nov-2000 Date2 18-Nov-2000. When I run the file I only get the first Target 1 (for the first example). It looks like it is only using the first date to determine when to break and not looking at both.

[Fangjun Jiang]

It is correct. I included your data in the code. Notice the last line returns 1.

[luke]

I stand corrected. I must have changed something in the code.

Thanks! Now I need to go thru this so that I understand how it works

Answer 2

[Nathan Greco]

Here's a dirty method, but no handwritten loops as requested.

Assume tmp contains your 22x3 cell array of data (the 4th column will be computed).

   %storing strings as unique identifiers
   tmp2 = cellfun(@(x,y)[x y],tmp(:,1),tmp(:,2),'un',0);
   %get unique blocks:
   [m m m] = unique(tmp2);
   %get error column:
   err = [tmp{:,3}];
   %get target:
   target = arrayfun(@(x)any(ismember(accumarray(m,err,[],@(x)min(x)),x)&x<.10),err,'un',0)';
   %add target to original cell array:
   tmp = [tmp target];

Hopefully this helps. It does the trick, but not sure about the speed. Anyone want to speed mine up?

Also note that loops don't necessarily mean SLOW anymore in Matlab. Try some timings on solutions provided versus your own to see what is best for you.