I am working on the Pareto model. After applying Reduced log likelihood for ECM algorithm , the MLE values of one of the parameters is negative. Can this be accepted?
In general, yes, a maximum likelihood estimate can be negative. For example, the maximum likelihood estimate of the mean of a normal distribution when the data is just the singleton vector (-1) is -1.