In this paper we have applied Poisson regression model for count data, Zeger (1988) model and Zeger and Qaqish (1988) model on rain fall data collected from Dhaka station metrological department of Bangladesh. Here the main response of interest is a count time series variable that is the number of rainy day in a month. We tried to model count data, the number of rainy days in a month, on humidity and temperature as the independent variable. We found that the iterative weighted least square estimation proposed by Zeger (1988) performs better that quasi likelihood estimation for dependent data. Also that Poisson regression model compared to Zeger (1988) model performs better in our case. Also the Markov regression approach, Zeger and Qaqish (1988) model performs better with full set of independent variable. One limitation is that no model is judged in real life situation by simulating data due to lack of adequate understanding and computation.