Is it a correct process of normalization to divide a data series by the maximum value of that series . Later these may be used for statistical analysis like pca
Use of just only maximum value is not a normalization, you will scale your series if you only use a single value.
Usually not. Extreme values are most unreliable (that is: the extreme value as a random variable representing a sample statistic has a relatively huge variance for most distribution models of the original variable, and often neither expectation nor variance of the extreme value distribution are defined).
For explorative purposes, often z-scores are used, sometimes of the log-transformed variables (if they are strictly positive and skewed).
I agree with Jochen. The problem is that people use the term "normalize" in many different ways. Best, David Booth
- Badges
- Science topic
- Similar topics
- Correction