Variance of sum of random variables independent when not consecutive

293 Views Asked by At

So I have 20 different 'weeks' that are being considered as part of a weight loss exercise. The weight lost each week is distributed normally around mean $\mu =0.8kg$ and std. deviation $\sigma^2= 0.1kg$.

The variance of the total will be the sum of the 20 variances, plus $2\cdot{{20}\choose {2}} \cdot \mathrm{Cov}$. With the correlation between weeks given as $-0.1$, the sum I have attempted as: $(20\cdot0.1^2) + (2\cdot{{20}\choose{2}})\cdot(-0.1\cdot\sqrt{0.1^2 \cdot 0.1^2})$ results in a negative variance of $-0.38$. Could someone please assist me as to where I have gone wrong?

Thanks very much

1

There are 1 best solutions below

1
On

Disclaimer: This is, so far, one of my most downvoted answers on the site. Since it is perfectly correct and answers the question as formulated at the time, the downvotes might be due to some extra-mathematical reasons. In any case... happy reading!

If the correlation between each pair of successive items amongst $n$ items with common variance $v$ is $c$ and if all the other correlations are zero then the variance $v_n$ of the sum of these $n$ items is $$v_n=nv+2(n-1)cv=(n+2c(n-1))\,v.$$ This is because the square of a sum $\sum\limits_{i=1}^nx_i$ of $n$ terms has $n$ terms $x_i^2$ and $2(n-1)$ terms $x_ix_j$ with $|i-j|=1$. Thus, the expectation of the sum of $n$ random variables in the present situation yields $n$ times their common variance plus $2(n-1)$ times the common covariance $cv$, all the other $(n-2)(n-1)$ covariances being zero.

In the present case, $n=20$, $c=-0.1$ and $v=0.01$ hence the variance of the sum is $$v_{20}=16.2\times0.01.$$ Note in addition that, by the computation above, admissible values of the correlation $c$ are such that $$c\geqslant-\frac{n}{2(n-1)}$$