Hi Sravani,

This sounds like a job for a maximum likelihood, or Bayesian, analysis: the first step is clarify the problem in order to write down a likelihood function, i.e., the probability of your observed data given the parameters of the broken power-law model. One point to clarify, is your fitting problem a regression (i.e., are you trying to fit a broken power-law shape as the mean of some data with noisy y-axis variable and known x-axis variable, perhaps?), or is it the shape of a distribution of objects (e.g. visualised as a histogram) which you aim to model with a broken power-law shape (having fixed limits and normalised appropriately)?
If you can write down the likelihood function the rest should be easy …

cheers,

Ewan.