Negative Values in Time Series

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

smurac1986
Posts: 12
Joined: Fri May 06, 2011 7:49 am

Negative Values in Time Series

Postby smurac1986 » Sun Jan 15, 2012 5:25 am

I have a series of CPI YOY% data which needs to be in logarithmic form. The problem is a small handful (maybe 6 or 7 out of the total 60) are negative values which obviously cannot be logged. I was wondering how this could be overcome or remedied on Eviews. Any feedback would be great.

Thanks,
Shane

startz
Non-normality and collinearity are NOT problems!
Posts: 3797
Joined: Wed Sep 17, 2008 2:25 pm

Re: Negative Values in Time Series

Postby startz » Sun Jan 15, 2012 7:56 am

Are you sure inflation needs to be logged? It's relatively unusual to log a percentage rate of change.

As you point out, you can't log a negative number. This has nothing to do with EViews.

Sometimes people add a small positive constant to the entire series before taking the log. At a mechanical level, that let's one proceed but it's not really the same.

loooooo
Posts: 11
Joined: Wed Dec 14, 2011 4:11 am

Re: Negative Values in Time Series

Postby loooooo » Mon Jan 16, 2012 12:23 am

Hi Shane,

Not sure if you tried to log a % rate of change, but if you did, it's wrong.
Since the rate of change is defined as "P(t)/P(t-1)-1," you should consider adding back 1 and log it, which will become "log(P(t))-log(P(t-1))" now.
Inflation in annual term can't be higher than 1, so converting to log rate like this won't yield any negative values I reckon.

Hope this helps.


Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 2 guests