Is Forecast Evaluation sensitive to negative values?

[sunnyeviews]

Dear Sir or Madam,

Is Forecast Evaluation sensitive to negative values? I am trying to run the Forecast Evaluation (simple mean, trimmed mean, simple median, least-sqaures, mean square error, and MSE ranks), but I receive an error message: positive or non-negative argument to function expected.

Yours faithfully,

[EViews Gareth]

Could you post your workfile?

[sunnyeviews]

attached. Thank you very much. :D

[EViews Gareth]

What exactly did you do?

[Oladapo]

Please house, I need help on how to forecast GDP data of my country, Nigeria, from 2014 to 2060. I already have data from 1960 to 2014 but I need forecast data till 2060.

[trubador]

Please house, I need help on how to forecast GDP data of my country, Nigeria, from 2014 to 2060. I already have data from 1960 to 2014 but I need forecast data till 2060.

You need to build a model to do that. I think you are working with annul data, which covers more than 50 years. Since you are interested in forecasting the next 50 years, you'll need a structural model. If you can find data on capital stock and labor force as well, then "growth accounting" approach would be a good place to start. I cannot think of any simple/easy alternative way other than univariate decomposing of the series through fitting a log-linear or polynomial functional form (you also can try ETS feature of EViews for that matter):

Code: Select all

equation eq.ls log(gdp) c @trend equation eq.ls log(gdp) c @trend @trend^2

Of course, you'll need to be careful about the breakpoints and structural changes in the data.