0

So I have a set of data, units purchased per month over a 4 or 5 year period. I have about 3000 different items, but here are 2 examples of what Ive done:

Item 1: Mean 24 STDEVA 18 It has purchases practically every single month of each year.

Item 2: Mean 12 STDEVA 15 It has at least 1 or 3 months with purchases of 0.

I do understand that stdev doesnt treat 0's differently than it does outliers. So for example, if most numbers are 1, then a 0 now and then wont increase your std dev. But if most numbers are 100, then a 0 now and then will increase it, right?

My question is, how can I determine which items are safer to purchase (inventory) such that I wont be so affected by large stdev? Iow, 18 of 24 is still a pretty big number. Its telling me that 68% of all values in that set are +/-18 units away from 24. That could be as low as 6 or as high as 42. On the other hand, 15 of 12 is just as bad.

How should I use the std dev to determine my safety level of purchases for a given year?

Set 1: 14,80,14,52,12,71,23,12,32,9,14,28,2,9,13,74,25,62,18,58,19,21,14,11,6,25,37,1,46,48,83,29,8,26,50,15,26,66,19,65,93,4,16,98,62,24,0,0,6,5,33,0,1,70,0

Set 2: 4,4,13,30,22,50,16,4,60,16,31,35,10,10,10,8,10,23,0,3,18,0,10,8,37,73,2,11,9,18,4,0,3,11,19,3,7,0,3,0,0,2,0,4,4,14,11,0,0,0,0,4,0,0,0

marciokoko
  • 131
  • 4

1 Answers1

2

The two data illustrations both have very large s.d.'s relative to the mean. Therefore the upper predicted limit for future data is very large. This single method of analysis does not seem to be adequate for your purpose. Perhaps the data needs to be tied to months of the year, or to your frequency of advertising, or to the seasons, or to whatever is driving the large variation.

daveM2
  • 21
  • 1