I have data compiled by someone else where score averages have been computed over time- averages range from 0-100. The original scores have negative values in many cases and the average would have been negative also, raw average ranges from -30 to 90. How is this 'normalization' accomplished?
Thanks
Asked
Active
Viewed 1,998 times
-3

Tailltu
- 59
- 1
- 4
-
Do you mean normalize to (0,1)? Please explain clearly what you are asking. – Rob Hyndman Aug 03 '10 at 01:08
-
This question is proposed to be closed. See this meta thread: http://meta.stats.stackexchange.com/questions/213/list-of-candidate-questions-to-be-closed – Aug 03 '10 at 08:58
2 Answers
4
Or more generically,
Index = ( ((RawValue - Min(Raw))/(Max(Raw)-Min(Raw)) * (Max(Out)-Min(Out) ) + Min(Out)
Where Raw
is the input vector, Out
is the output vector, and RawValue
is the value in question. Srikant's answer is the same for an output range of 0 to 100.
To convert back, rearrange to get:
RawValue = (((Index - Min(Out))*(Max(Raw)-Min(Raw)) / (Max(Out)-Min(Out)) + Min(Raw)

RobinGower
- 141
- 2
1
The question is a bit unclear. But, perhaps the normalization you are looking for is this:
Normalized Score = 100 * (Raw Score - min(Raw_score)) / (max(Raw Score) - min(Raw Score))