davidsmith73
Graduate Poster
- Joined
- Jul 25, 2001
- Messages
- 1,697
No, or more accurately, "it depends". Here's a simple example of how it might pan out:
In a binary data point, the result might be 1 if ANY observation of a phenomenon is made, 0 otherwise. In a variant data point, that may be a grade of how "big" that phenomenon might be, from 0 (not seen) to 9 (huge). So ANY value over zero could be a taken as positive result in a binary measuring system. However the totality of the results in the distributive scoring method might be that they were all usually zero, sometimes 1 and occasionally 2. That is, the scores are a really small fraction of the potential values. However, translating to binary (zero/non-zero) will artifically inflate the results.
Let's try this with 10 made-up scores, from a range of 0-9 each.
1 0 1 2 1 0 0 1 2 1; Average = 0.09, stddev = 0.0737864787
Let's convert these to zero/non-zero binary, range 0-1.
1 0 1 1 1 0 0 1 1 1; Average = 0.35, stddev = 0.241522946
Nearly four times larger "positive" result = artifical inflation. (Thanks, Excel!)
I don't get this. If your method is valid, why didn't the data collected using the distributed or FIDO method and then converted to binary scores inflate the results? (Tables 4 and 5)