none
Generalized least Square Model for Differenced Interrupted Time Series Analysis RRS feed

  • Question

  • I carried out a Durbin-Watson test with a maximum lag of 13 after carrying out preliminary linear regression model for interrupted time series data and I did not get any significant lags.

    I also plotted the acf and pacf of the residuals of the model and none of the spikes are beyond the threshold. Therefore, I concluded that the data are not serially correlated. However, I tested the outcome variable for trend and drift and I had found that there are both a monotonic trend and a drift.

    So, the data are non-stationary. I wanted to carry out the ARIMA model to remove the trend and seasonality but there is no corARIMA model in gls function.

    Should I run linear regression by ignoring the trend and seasonality or how can I find the corARIMA model?




    • Edited by MelakuH Tuesday, April 16, 2019 2:46 PM
    Tuesday, April 16, 2019 1:43 PM

All replies

  • Just a comment:

    When using Grammarly, then you need to correct or ignore all marked words. Otherwise you're posting the Grammarly HTML, which makes posts unreadable.

    Tuesday, April 16, 2019 2:00 PM
  • Thank you. I have edited the content now

    MelakuH


    • Edited by MelakuH Tuesday, April 16, 2019 2:48 PM
    Tuesday, April 16, 2019 2:47 PM