Find the distribution of $W= \frac{X_1+X_2}{|X_1-X_2|}$ , $X_i = N(0,\sigma^2).$
Moreover: What hypothesis testing can be performed using the statistic in Q1 when is unknown?
My attempt:
I know, by using moment generating functions and using any stats textbook that $X_1+X_2$ has the distribution $N(0,2\sigma^2)$.
However this is where I am stuck.
I know that $X_1-X_2$ has the distribution $N(0,2\sigma^2)$ however I am unsure whether the absolute value makes a difference. If it doesn't the Wolfram Alpha page on Cauchy Distributions says that the ratio of two random normal variables centered at 0 is Cauchy. However I am not sure whether this is the case, and if it is the case, what assumptions can be made about the hypothesis testing. Any clarification would be great.