for (i in 1:n) { tmp = ugarchroll(spec[[i]], R, n.ahead = 1, forecast.length = 1500, refit.every = 50, refit.window = 'moving', windows.size = 1500, solver = 'hybrid', calculate.VaR = FALSE, cluster = cluster, keep.coef = FALSE) while(!is.null(tmp@model$noncidx)){ tmp = resume(tmp, solver = 'gosolnp', fit.control = list(scale = 1), solver.control = list(tol = 1e-07, delta = 1e-06), cluster = cluster) } fitlist[[i]] = as.data.frame(tmp, which = 'density') print(i) }]]>

Here’s the code to replicate the qq plot (which you can also call by type ‘plot(fit, which=9)’):

vmodel = fit@model$modeldesc$vmodel zseries = as.numeric(residuals(fit, standardize=TRUE)) distribution = fit@model$modeldesc$distribution idx = fit@model$pidx pars = fit@fit$ipars[,1] skew = pars[idx["skew",1]] shape = pars[idx["shape",1]] if(distribution == "ghst") ghlambda = -shape/2 else ghlambda = pars[idx["ghlambda",1]] rugarch:::.qqDist(y = zseries, dist = distribution, lambda = ghlambda, skew = skew, shape = shape)

The main functionality can be found in the ‘rugarch:::.qqDist’ function.

]]>Are the instructions on the r-downloads page not clear?

https://www.unstarched.net/r-downloads/

And how is this different from what the function allows you to do?

You choose the p-value for the cutoff, the coefficients above that are dropped and the model re-estimated. You can then repeat the process by re-submitting the object to the function.

Not sure I understand what you mean “use them in a new dataset”.

Have you tried using the “rcov” and “sigma” methods on the estimated objects? Have a look in the rmgarch.tests folder in the src distribution for examples.

Also, make sure you download latest version which contains some fixes to the dcc simulation from the bitbucket repository (see : R-downloads).

Alexios

]]>Best regards

Jostein Aaland

Hi, the model has a full set of methods to work with (specification, estimation, filtering, forecasting, simulation, rolling re-estimation/forecast and the usual extractor methods: fitted, sigma, quantile, pit etc). However, you’ll need to familiarize yourself with R somewhat (BTW: R studio is an R gui…you don’t need it to use R, though it is very good), as well as the use of xts formatted data and R’s time/date functions in order to understand how to format the input data for the mcs model.

]]>