0
$\begingroup$

I have this part of a project which states this: Once you choose the best three models for each series (according to AIC, the PACF and ACF, and "from general to specific), the next step is to estimate these 6 models. You should present the final models (estimates and t-stats) and the proof that the residuals are white noise (you can use for this Q-stat or check if residuals have serial correlation with 5% significance level). In the previous step I had to find the parameters of the AR model, and in order to check if residuals are white noise I run the Ljung-Box test in this way (given the cutoff of the PACF of my differentiated time series is 6, I wanted to validate an AR(6). I defined x as my diff(log(gdptimeseries)). ar6=arima(x,order=c(6,0,0)

Call:
arima(x = gdplogdiff, order = c(6, 0, 0))

Coefficients:
         ar1      ar2     ar3      ar4      ar5     ar6  intercept
      3.3058  -4.5471  3.0440  -0.6789  -0.3044  0.1651      0e+00
s.e.  0.0579   0.2025  0.3345   0.3343   0.2021  0.0576      4e-04

sigma^2 estimated as 8.669e-09:  log likelihood = 2250.22,  aic = -4484.44

What I'm I now supposed to do? how do I estimate the parameters and the t-stats of the model?

$\endgroup$

1 Answer 1

2
$\begingroup$

The estimated values of the autoregressive parameters are already reported in the output you provided:

Coefficients:

         ar1      ar2     ar3      ar4      ar5     ar6  intercept
      3.3058  -4.5471  3.0440  -0.6789  -0.3044  0.1651      0e+00

The associated standard errors are listed immediately below these estimated values:

         ar1      ar2     ar3      ar4      ar5     ar6  intercept

s.e.  0.0579   0.2025  0.3345   0.3343   0.2021  0.0576      4e-04

The t-values are obtained as the ratio of the estimated autoregressive coefficient to its corresponding standard error. For example, the t-value for testing the null hypothesis thar the (true) first autoregressive coefficient is 0 against the alternative hypothesis that it is different from 0 is given by:

t = 3.3058/0.0579

and you can compare its absolute value against the critical value 1.96 from a standard normal distribution to decide whether you can reject the null hypothesis in favour of the alternative hypothesis based on the evidence in your data at the usual 5% significance level.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.