Let $x_1,...,x_n$ be a random sample from some population with $n\geq3$ and with at most $n-2$ sample points being equal. How can one show $$ \sum_1^n (x_i-\bar{x})^2 = \sum_1^{n-1} (x_i -\bar{x}^*)^2+\frac n{n-1}(x_n-\bar{x})^2, $$ where $ \bar{x}=\frac1n\sum_1^nx_i$ and $ \bar{x}^* = \frac 1{n-1} \sum_1^{n-1} x_i.$ This is a useful identity since it would allow me to have an upper bound on the deviation of $x_i$ from the mean. That is $$ \max_{1\leq i \leq n} | x_i - \bar{x}| < \frac {n-1} {\sqrt{n}} S, $$ where $ S^2 = \frac 1 {n-1} \sum_1^n (x_i - \bar{x})^2.$
Thank you very much for your help.
Source of the question
A Matrix Formulation on How Deviant an Observation Can Be
Author: Ingram Olkin.
Source: The American Statistician, Vol. 46, No. 3 (Aug., 1992), pp. 205-209
https://www.jstor.org/stable/2685215
I will add shortly what I've tried so far.