Data Analysis Based on Multivariate Time Series by States
I Gusti Ngurah Agung · 2013
Every analyst should use his/her best subjective judgment to select sub-sets of exogenous variables by levels of their importance. So data analysis should be done using multistage regression analyses by inserting the sub-sets of the variables one-by-one. Every researcher should present statistical results based on several regression models in all data analyses, beside descriptive statistical summaries. Corresponding to this idea, separate sections of this chapter present illustrative examples based on simple models using only two exogenous time series, namely (X1_i,X2_i) = X_i, and an endogenous time series, namely Y1_i, for a small number of states (N). This is extended to X_i = (X1_i,X2_i,X3_i) and Y_i = (Y1_i,Y2_i), as the base models. The models discussed are models based on (X_i,Y_i,Z_i) for independent states, models based on (X_i,Y_i,Z_i) for correlated states, simultaneous SCMs with trend, models based on (X1_i,X2_i,X3_i,Y1_i,Y2_i) for independent states, interaction models, and discontinuous time-series models.