I have one time series variable, and i divide the time period into different sections according to different governments (all sections have unequal number of years). Now, i wish to compare the variations in different time sections by using suitable time series model.
As your sample has unequal size, you can try non-parametric tests which is equivalent to paired sample t-test.
Mann- whitney U test is the widely used test in this case.
you can apply this test in SPSS, PSPP like statistical softwares
mann-whitney test is used when there are two samples, and for the case of more then two samples we have kruskal walis H test. and noth of these tests are utilized for the testing of location parameter not for variation. but i want to compare the variation of all these periods by utilizing proper time series model. kindly guide me in this regard.
I don't agree it is as simple as that.These Mann-Whitney and Kriskal-Wallis tests assume i.i.d.observations. Time series data are realizations of independent random variables. One partial solution can be to use ARIMA modelling with so-called interventions, due to Box and Tiao (1975), "Intervention Analysis with Applications to Economic and Environmental Problems", Journal of the American Statistical Association, Vol. 70, No. 349, pp. 70-79. Most implementations will only consider changes in levels and assume the same stochastic process except for that. I have something for scatter variations. I think you should show your data for better advice.
If I understood you correctly, you want to compare different policy regimes. Why not try dummy variables? If you have, say three regimes, construct a model with two dummy variables (the third will be the reference category). You then run OLS and test the significance of the coefficients attached to the dummies.
May be it can help you Mr Asad..
https://en.m.wikipedia.org/wiki/Chow_test