1

I was attempting to replicate the results of a paper and ran PACF plots through R Studio, Stata, and EVIEWS for the UK debt-to-GDP ratio. I received quite different plots in return. Can anyone shed some light on why this is the case? My understanding is that the regressions that make up the PACF should return the same results regardless of platform.

Figures from RATS, RSTUDIO, EVIEWS and STATA respectively below:

enter image description here

enter image description here

enter image description here

enter image description here

steve
  • 107
  • 1
    Were all three different, or were two of the three the same and one different? – jbowman Jul 21 '23 at 17:50
  • All three were different. And all three were different from the paper I was trying to replicate. I have the same data and the time series itself is identical when plotted. The paper used RATS I believe. – steve Jul 22 '23 at 18:52
  • I am now upvoting your question, as they really should be the same, or at worst only slightly different! – jbowman Jul 22 '23 at 18:53
  • Thanks. I will try to upload plots on Monday. – steve Jul 22 '23 at 18:58
  • EDIT: Now that I look again EVIEWS and R are producing the same plot. I checked the actual correlations and they are identical. So, three different plots from four different packages. – steve Jul 24 '23 at 10:43
  • 1
    This question is confusing because the PACF and ACF plots should differ and you haven't even labeled or described all the plots. Moreover, because there are different estimators of the ACF (see https://stats.stackexchange.com/questions/81754 for some discussion), there's no reason to suppose either plot should be identical among software packages: the first thing to do is consult the documentation so you know what your software is actually doing! – whuber Jul 24 '23 at 14:00
  • Pardon me I see what you mean. They are all PACF plots. I'll edit the question. – steve Jul 24 '23 at 15:22

0 Answers0