Apply logical matrix to dataset

I have a logical matrix (IdxReturns) with the dimension (589x693) which I would like to apply when calculating the product of my return series (MyReturns). However, when I attempt to apply the logical matrix I get a 9387x1 matrix. How do I go about getting around this? I am using a command that looks like the following:

TempReturn = prod(MyReturns(IdxReturns),2);

Because I am calculating the product function across the rows, I would expect a 589x1 matrix as a result. What am I doing wrong? Do I need to apply the logicals separately?

Answer 1

[Brian]

Thank you guys for your help with this. I actually ended up finding two different solutions that would work. If I apply the logical matrix to the 589x693 matrix I can simply apply a reshape function to get it back to the intended dimensions. Also, if I apply the logical arrays separate from one another rather than creating a logical matrix first, the TempReturns dataset keeps it's form appropriately. I chose this solution and the ending code format can be seen below.

TempReturn = prod(MyReturns(IdxTickers,IdxDates),2)

Answer 2

[Fangjun Jiang]

Try this example to see how logical index works.

a=magic(3);
b=logical([1 0 1;0 1 0; 0 0 1]);
c=a(b)

It's picking the element from matrix a according to the logical index b and put the result in a vector.

To make your code work, you need to do the following.

MyReturns(~IdexReturns)=1
TempReturn = prod(MyReturns,2);

It sets the value of the non-selected elements to be 1, because you are doing a prod() operation.

Answer 3

[Walter Roberson]

t = MyReturns;
t(IdxReturns) = 1;
TempReturn = prod(t, 2);

[Walter Roberson]

It is not clear that your logical matrix corresponds to selecting a mix of rows and columns, so it is not clear to us that a rectangular output would even be possible.

Try

TempReturn = prod( MyReturns(find(any(IdxReturns,2)), find(any(IdxReturns,1))), 2)

If that does what you want then probably using a logical matrix is not the best approach for your needs.

If you really need a logical matrix...

TempReturn = prod( reshape(MyReturns(IdxReturns), sum(IdxReturns,2), sum(IdxReturns,1)), 2);

and if that bombs out with a complaint about reshape changing the number of elements then the implication is that your IdxReturns does not mask out row / column combinations.

Answer 4

[Brian]

Thank you guys both for your answers to this point. With my logical matrix I am not only limiting along the x axis but the y axis as well. In other words, my original matrix has 593 Funds down the y axis but I may be limiting that universe of funds to 500. Therfore, I need to find a way to exclude the 93 as opposed to setting them = 1. Otherwise when I calculate prod(MyReturns,2) I will get a product for rows that I was attempting to exclude.

Thanks for your help

[Fangjun Jiang]

How do you dump those 93 funds? If no criteria, it's easy to do.

a=rand(593,3);

b=a(1:500,:)

Answer 5

[Brian]

I wish it were that easy. I've got plenty of other criteria that is helping me to determine whether a fund is in or out of the calculation for this given period. Maybe this illustration will help.

I import a file that is broken into 3 parts (Horizontal header - I create IdxDates) and (Vertical header - I create IdxTickers) and then the 589x693 matrix of fund level returns. Maybe I should avoid using the repeat matrix and somehow filter my dataset vertically and horizontally before doing my prod() calculations.

 %Determine my list of funds from the original 593
  IdxTickers = ismember(MyTickers,PGHoldings(IdxPGFilt & IdxPGDateFilt));
 %Filter my dates for the desired time period
  IdxDates=MyNumDates>=BegDate & MyNumDates<=EndDate;
 %Use repeat matrix to give me a 589x693 logical matrix
   IdxReturns = repmat(IdxTickers,1,DateCnt) & repmat(IdxDates,TickCnt,1);
 %Calculate returns (this is where I'm having the issue)
   TempReturns = prod(MyReturns(IdxReturns),2);

Thanks again - I really appreciate it.

[Fangjun Jiang]

If IdxReturns is the correct index, why this code won't work? The size of matrix is dependent. There is nothing here that can guarantee the size to be 500, or 589.

MyReturns(~IdexReturns)=1

TempReturn = prod(MyReturns,2);

[Brian]

Maybe that's where I'm going wrong. I need to guarantee that the size is 500 if that's the number of funds that needs to be in the calculation. Maybe I take "MyReturns" and reduce it to the 500 and then take "MyReturns" and shrink it by the IdxDates before doing the prod() calculation. I think what you're telling me is that I can't apply a 589x693 logical to a 589x693 matrix, use the product function across the rows, and expect to get either 500 or 589 results back.