more efficient way of checking if conditions are satisfied?

Hi, i have a stimulus that can change in three variables, in any combination (e.g no change, 1 change, 2 changes or all three changes co-occur) and for each trial i want to categorise the change that occurred for later recording in an excel spreadsheet.

my code, which currently works fine, seems too long and a bit like overkill. I was just wondering if anyone could suggest a more efficient method of checking these statements?

thanks

code below-

        if orientation_seq{k} == phi && contrast_seq{k} == C && ram_seq{k} == A;
            num_changes = 0; type_changes = 'no_change'; co_occur = 'no';
        elseif orientation_seq{k} ~= phi && contrast_seq{k} == C && ram_seq{k} == A;
            num_changes = 1; type_changes = 'orientation'; co_occur = 'no';
        elseif orientation_seq{k} == phi && contrast_seq{k} ~= C && ram_seq{k} == A;
            num_changes = 1; type_changes = 'contrast'; co_occur = 'no';
        elseif orientation_seq{k} == phi && contrast_seq{k} == C && ram_seq{k} ~= A;
            num_changes = 1; type_changes = 'RAM'; co_occur = 'no';
        elseif orientation_seq{k} ~= phi && contrast_seq{k} ~= C && ram_seq{k} == A;
            num_changes = 2; type_changes = 'orientation & contrast'; co_occur = 'yes';
        elseif orientation_seq{k} ~= phi && contrast_seq{k} == C && ram_seq{k} ~= A;
            num_changes = 2; type_changes = 'orientation & RAM'; co_occur = 'yes';
        elseif orientation_seq{k} == phi && contrast_seq{k} ~= C && ram_seq{k} ~= A;
            num_changes = 2; type_changes = 'contrast & RAM'; co_occur = 'yes';
        elseif orientation_seq{k} ~= phi && contrast_seq{k} ~= C && ram_seq{k} ~= A;
            num_changes = 3; type_changes = 'orientation & contrast & RAM'; co_occur = 'yes';
        end

Answer 1

[dpb]

Try something like

concur={'no' 'yes};                                  % table of concurrence answers
ref=[phi C A];                                       % vector of reference conditions
inp=[orientation_seq{k} contrast_seq{k} ram_seq{k}]; % and the inputs less verbosely
changes=[inp-ref]~=0;                                % logical of nonzero locations
num_changes=sum(changes);                            % how many
co_occur = concur((num_changes>1)+1);                % concurrent?
...

Store your other conditions in other lookup tables and build the other variables based on position in the changes vector.

ADDENDUM

The prime advantage of the above is that it is much more easily adapted to changing the number of inputs; only the data tables need changing, basically.

ADDENDUM(2)

Got to thinking that the latter above suggestion isn't necessarily all that clear as to how to implement...so the idea is as follows...

>> types={'orientation' 'contrast' 'RAM'};    % the non-zero types array

Having computed the differences as above so have the changes logical vector and number, then for a sample case where

>> changes=[1 1 0]~=0;  % a test vector where first two are changed...
>> num_changes=sum(changes); % recompute number 
>> fmt=[repmat('%s & ',1,num_changes-1) '%s'];  % dynamic format string
>> type_changes = sprintf(fmt,types{changes})   % build the variable
   type_changes =
   orientation & contrast
>>

The dynamic format string to concatenate the pieces with the desired punctuation between and the use of the curlies "[]" on the array subscript inside the sprintf call which creates a comma-separated list that can be handled as an input vector are the two tricky parts...

Note also that the above is done for the case of having at least one change; as IA suggested, preset the defaults for no changes then after compute changes vector and number then just an

if num_changes 
  ...
end

block for the above for the two compound variables is all that's needed.

[kwn500]

hi, thanks so much for this. this is the sort of thing i was looking for.

i don't have too much experience with coding, and i'm unsure of how to fit this into a loop concisely (i'm sure it's very obvious).

i understand that i require: if num_changes = 1/2/3 clauses, but how do i specify if the the specific change, rather than, for example;

if num_changes = 2;
    if changes == [0 1 1];
    do something
    if changes == [1 0 1];
etc.

i'm getting a little bit confused- i don't think the loop can distinguish between the different changes without these explicit clauses (if i am right?)

but thank you so much, this is very helpful

[dpb]

"... i require: if num_changes = 1/2/3 clauses..."

No, then you definitely don't understand. :)

That's what the vector solution eliminates for you as demonstrated for the case of two changes--note the variable contains both elements of the pattern array with only a single statement formatting the string into the variable. It'd done the same with 1 or 2 or 3 True locations; I just chose two as the demo.

The changes vector is defined as I showed in the first posting; the case I used in the second was just an example of what the result could be to use in the lookup table.

changes=[inp-ref]~=0;

will return a logical vector which locates which of the N=3 variables are changed (or none, of course, if all are 0 difference). Using that vector in logical indexing expressions gets everything in one reference, which is the power of Matlab vectorized syntax; the whole reason, in fact, for "MATrix" in the name.

The only place where some loop might come into play is that your initial line has a subscript k on these variables; if there are multiple instances then that would have to be dealt with, but NOT any loop for the number of elements that change for a given set; that's automagic if you'll try it as written.

[kwn500]

sorry, this took me a little while to get my head around, but i think i've got it. :) it works perfectly and was exactly what i needed- thanks so much again!

[dpb]

"... works perfectly and was exactly what i needed-..."

So Accept the answer to at least indicate the thread is closed if nothing else...

Answer 2

[Image Analyst]

You could shorten it a bit by defining defaults in advance

num_changes = 0; 
type_changes = 'no_change'; 
co_occur = 'no';

Then getting rid of the first if block and setting those values in the other if blocks only if they are different than the defaults.