If 'msgdata' and 'moddata' are FLOATs, then 'result' must be a FLOAT and there should be no FLOAT( ) conversion call necessary in this example. That is probably besides the point for your question, however. Perhaps if you called "help, result" instead of "print, result", the root of the problem would be clearer. REGRESS always returns an array. I am guessing that in your example, it is returning a 1-element array. In IDL a 1-element array multiplied by a multi-element array of any dimension will return a 1-element result - it will be the product of the first element of the multi-element array times the value of the 1-element array, one single value. That is very different from the result one gets if one multiplies an array by a scalar. When a programmer "manually types a value" he/she is usually typing a scalar. That would explain the differing results you are getting from your variable and your manual typing. Now compare the result of: outdata[temp] = msgdata[temp] * result to the result of: outdata[temp] = msgdata[temp] * result[0] ; The subscript converts this to a scalar I bet that fixes at least part of your problem. You may have another problem here, though, as well. What is the meaning of this? temp = where(moddata ne 0 or msgdata ne 0, count) 'temp' in this case ends up with subscripts valid for 'moddata' mixed with subscripts valid for 'msgdata'. Is this really what you want. In the example you showed above, I suspect that your call: outdata[temp]=msgdata[temp]*float(result) had a subscript in the location temp[0] that was non-zero in 'moddata', but was zero in 'msgdata'. That is why you ended up with all those 0 values in your 'outdata'. James Jones