I have gone through the models designed by Mr. Vinay Kshirsagar & Mr. Omkar Kshirsagar. The Models have been built by taking into consideration of the data on various Economic parameters for India and USA of last 36 years from 1980 to 2016.
The said input data used for these models is not tested for or validated by me on the assumption that the data taken is correct, unbiased and satisfies the condition of uniformity.
The models have been worked out by using Mathematical method, Exponential smoothing average method and Moving Average method. For exponential smoothing method the value of smoothing constant is taken as 0.7 for this value logical explanation is given. Even in Moving average method weights are 0.55, 0.3, 0.1 and 0.05 which represent the proportion of the impact.
All the 3 models are tested for goodness of fit. The method used is regression analysis. This confirms the applicability of the model. The results show- Multiple correlation over 90%, dependency over 85%, significantly lower standard error, residual graph having random pattern, Significant F value ( P-value of overall model is less than 0.05 ). This shows all values are well within the acceptable limits which confirm the overall work-ability of the models to forecast the exchange rate.
Although some coefficients individually doesn't properly indicate the changes in exchange rate however all the coefficients together greatly influences the movement of exchange rate with very high degree of accuracy.
Also considering that the future prediction can never be 100% accurate however this Model is tested with past data for its fitness. The actual results are compared for the Financial years 2016 and 2017. Only the time will confirm its accuracy in the years to come.
I understand from authors that major Economic variables which impact the exchange rate are considered in building the model.
Further when asked about the removal of variables having insignificant impact on the model on the principle of p value and multicollinearity, it has been clarified that the author has done extensive research and evolve this model which has high degree of correlation and dependency. If one of the variables is removed in long runs it impacts. The final combination of these 7 variables gives the best results.
We wish authors all the very best in their future endeavours.
Mrs. Milan Gholba
Retd. Associate Professor,
Department of Statistics,
VPM’s B. N. Bandodkar College of Science, Thane
Dated: 28th September 2017