1

I have an experiment which evaluates decision making processes while buying a given product (wine).

I use MouseLab Web to gather the data. Basically the user has multiple attributes, that describe his product: Price, Brand, Type, Country of origin

The user can open up the different Attributes (displayed as cards) and have a look at the values. When he opens an attribute, all others are closed. The user may reopen any attribute as often as he wants.

That gives me an example output for one user like the following:

timestamp | attribute

like:

00001 | Type
00111 | Price
01111 | Brand
02222 | Price
03333 | Country
04444 | Brand

I am trying to evalute the data with a weighted ranking scale:

Type: 6
Price: 5
Brand: 4
Price: 3
Country: 2
Brand: 1

But this way I got the problem that a user that checks a lot of cards will have a huge score for each category comparing to a user that made his decision after only a few cards, so I can't compare the two users.

How would you normalize the ranking data to make it comparable?

user124531
  • 11
  • I'm not sure to understand the meaning of your weighted index. However, if you just want to (weightedly) count how many cards your used has checked, you may count every card just one time, in spite of having been checked more than once. A no-so-extreme approach might be to transform the number of times, for example by taking the square root. – Pere Jul 26 '16 at 22:25
  • the user must be able to check attributes more than once. Imaging standing in front of a board filled with wines. You would definately check attributes like type, price etc more than once until you make your decision. – user124531 Jul 27 '16 at 08:03
  • I understand that the user can check how many attributes as he wishes and how many times as he wishes. What I don't understand is what you want to do with your weighted index. Why do you need an index that values three times more if the user checks type than if he checks country? That's what I'm trying to understand before trying to give an answer. – Pere Jul 27 '16 at 10:08
  • @Pere thanks! Now I got it. I tried to start evaluating the data with a simple inversed ranking score = (number of decisions + 1) - current decision position since I had no better idea.. – user124531 Jul 29 '16 at 21:36

0 Answers0