Suppose $\alpha_1, ..., \alpha_n $ are independent identically distributed random variables, $ a_1, ..., a_n,b_1,...,b_n $ are non-zero constants. Is it true that if $ \sum_{i=1}^{n}a_i\alpha_i $ and $\sum_{i=1}^{n}b_i\alpha_i$ are independent, then $\alpha_1, ..., \alpha_n $ are normal variables?
-
1$\begingroup$ Can anyone explain me why was this question closed? $\endgroup$– Alexandre EremenkoApr 23, 2017 at 15:48
-
1$\begingroup$ It wasn't even a question. $\endgroup$– Franz LemmermeyerApr 23, 2017 at 19:33
1 Answer
This is called the Darmois-Skitovich theorem. Of course, one needs to add the condition that $a_jb_j\neq 0$.
The reference is MR0346969 Kagan, A. M.; Linnik, Yu. V.; Rao, C. R., Characterization problems in mathematical statistics. Translated from the Russian by B. Ramachandran. John Wiley & Sons, New York-London-Sydney, 1973.