In case of serial correlation, can we use AR model or it’s simply subjected to Dicky fuller test
Although my doubt is quite related to the los n – point 8 where sir explain for AR b/w 2 time series,
Which I am thinking primarily that if both rm and ri are forming a characteristic line with ols, then y can’t we make a simple regression only instead of going with df test, if both f and t test pass.
Haa, If there is serial correlation between them, then we can jump on AR subject to df test – and if not then co-integrated test if all not satisfy df test
And one more point – like evn in this the problem of unit root may arise, tht lead the series to be a random walk, which can be solve by 1st differencing
May be I am getting confused in terms, but Sir can you please give a bit clarity
In the case of regressing one-time series over the other ( say Rm and Ri), we can do so, only if both the time series are covariance stationary ( no unit root issue; you may check for the same using the DF test) or nonstationary but are cointegrated. (you may check for the same using the DFEG test).