3
$\begingroup$

We use turbidity to estimate suspended-sediment concentration (SSC)- our data was serially correlated. We ran an ARMA process and ended up with a AR (2) model. Our equation in log form is:

estlogi(SSC) = 1.35149 + 0.785099logi(Turbidity) - 0.00095 + 0.7492ei-1+ 0.2052ei-2

How can we input the ei-1 & ei-2 values in real time, that is, how do we enter residuals for current or future values when our turbidity/SSC data pairs are in the past (2010 to 2011). Someone did suggest an R script that will estimate K steps ahead for a specified ARMA model and past data- does anyone know this script?

Improve this question
$\endgroup$
4
  • $\begingroup$ Various software packages provide a machine readable file which could be easily converted to an r script. This file is based upon the model that was developed but is generic nature and useful to cover all univariate and causal time series models. You can approach SAS/SPSS or AUTOBOX and ask them to provide you with an example of this useful output. $\endgroup$
    – IrishStat
    Commented Sep 17, 2013 at 23:01
  • $\begingroup$ You say the model is an AR(2) but you talk about entering residuals? What precisely is e1-1 ? Is that SSC in the prior period ? . It is possible to restate your Transfer Function (armax model) BUT I would have to have the actual data that you used to form this model . Are all variables in log form ? $\endgroup$
    – IrishStat
    Commented Dec 17, 2013 at 18:26
  • $\begingroup$ Your equation does not describe ARMA model. You have a regression model with AR(p) disturbances. This means that for forecasting you can employ best linear unbiased prediction for GLS. This answer might help: stats.stackexchange.com/a/14442/2116 $\endgroup$
    – mpiktas
    Commented Oct 20, 2015 at 5:56
  • 1
    $\begingroup$ I'm not convinced this question should be closed as off-topic - there seems to be a statistical issue at the heart of this, it isn't just a request for code. $\endgroup$
    – Silverfish
    Commented Jan 28, 2016 at 11:35

0

Reset to default

Your Answer

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

Browse other questions tagged or ask your own question.