I need to do normalization and standardization both before my statistical analysis? or one of them? I would like to use some multivariate analysis for my data. My data was not normally distributed, therefore i did log-transformation for normal distribution. But still i need to do standardization? if it is i should use my original data for it or log-transformed data? 
 A: You need to do the right thing.


*

*there are data sets where it is the right thing to normalize

*there are data sets where it is the right thing to standardize

*there are data sets where it is the right thing to rotate via PCA

*there are data sets where it is the right thing to whiten using PCA

*there are data sets where it is the right thing to whiten and keep only the top components

*there are data sets where it is the right thing to do different things on different attributes (very common)

*there are data sets where it is the right thing to not do any of the above


The bad news is: there is no "if this then that" method to figure this out. You need experience, you need to understand the mathematical foundations and consequences. And you need to know your data. Since we don't have your data, we don't know what is right or wrong.
A: Another way to approach it is to read up on the estimator you are going to use. What are its assumptions?
And if you can and don't know better just test out all of the transformation possibilities with proper crossvalidation. 
Maybe you will succeed getting a good score and that's enough theory. 
