Is there a method to form a CI based on an MLE which is not the mean I know how to find confidence intervals for parameters based on sample means. I want to know whether there exists a method for finding confidence intervals based on MLE's which are not sample means. 
Any illustration would be helpful. Or, you may link me to some relevant website.
 A: Yes, there are several. I'll illustrate one. This doesn't always work (you may not be able to find a suitable quantity, for example).
One common approach to generate confidence intervals for parameters is via pivotal quantities.
A pivotal quantity is a function of the parameter and a statistic whose distribution doesn't depend on the parameter. You can then form an interval for the pivotal quantity (which will be the same no matter what the parameter value is), and from that back out an interval for the parameter.
So for example, consider the variance parameter for a normal distribution; $Q=(n-1)\frac{s^2}{\sigma^2}\sim\chi^2_{n-1}$.
To be specific, let's say n=15, and $s^2=28.5$.
From the $\chi^2_{14}$ distribution, a 99% interval for $Q$ is $4.075<Q<31.32\,$. So
$4.075<(n-1)\frac{s^2}{\sigma^2}<31.32$
$4.075<14\times 28.5/{\sigma^2}<31.32$
$1/31.32<\frac{\sigma^2}{14\times 28.5}<1/4.075$
$\frac{14\times 28.5}{31.32}<\sigma^2<\frac{14\times 28.5}{4.075}$
$12.74<\sigma^2<97.91$
So a 99% interval for $\sigma^2$ is $(12.74, 97.91)$
