R rugarch: $ operator invalid for atomic vectors?

Question

I am trying to make a huge nested for loop (optimizations be left for later) to fit all of the GARCH models available from rugarch.

This is my MWE that reproduces the error:

library(rugarch)

## Small parameter space to search over   
AR_terms = c(0,1,2)
I_terms = c(0,1)
MA_terms = c(0,1,2)

garch_p_terms = c(0,1,2)
garch_q_terms = c(0,1,2)

## Models to search over
var_models = c("sGARCH","eGARCH","gjrGARCH","apARCH","iGARCH","csGARCH")

for (x in var_models) {

    if (x == 'fGARCH') {

        for (y in sub_var_models) {

            for (AR in AR_terms) {
                for (MA in MA_terms) {
                    for (I in I_terms) {
                        for (p in garch_p_terms) {
                            for (q in garch_q_terms) {

                                cat(y)

                                spec = spec_creator('fGARCH', y, MA, AR, I, p, q)
                                garch = ugarchfit(spec = spec, data = apple['A'], solver = 'hybrid', solver.control = list(trace=0))

                                cat('Fit Success')

                            }
                        }
                    }
                }
            }

        }

    next ## To skip evaluating fGARCH as its own model with not submodel below.

    }    

    for (AR in AR_terms) {
        for (MA in MA_terms) {
            for (I in I_terms) {
                 for (p in garch_p_terms) {
                    for (q in garch_q_terms) {

                        cat(x)

                        spec = spec_creator(x, 'null', MA, AR, I, p, q)
                        garch = ugarchfit(spec = spec, data = apple['A'], solver = 'hybrid', solver.control = list(trace=0))

                        cat('Fit Success')    

                    }
                }
            }
        }
    }


}

)

with my spec_creator function defined here: (the fGARCH model allows a submodel family, which is the reason for most of the redundant code)

## Function to create the specs, purely to make the for loop area more readable.
spec_creator = function(model, sub_model, AR_term, I_term, MA_term, garch_p_term, garch_q_term) {

    require(rugarch)

    if (sub_model == 'null') {   
        spec = ugarchspec(variance.model = list(model = model, 
                                        garchOrder = c(garch_p_term, garch_q_term), 
                                        submodel = NULL, 
                                        external.regressors = NULL, 
                                        variance.targeting = FALSE), 

                          mean.model = list(armaOrder = c(AR_term, I_term, MA_term)))
    }

    else {
        spec = ugarchspec(variance.model = list(model = 'fGARCH', 
                                        garchOrder = c(garch_p_term, garch_q_term), 
                                        submodel = sub_model, 
                                        external.regressors = sub_model, 
                                        variance.targeting = FALSE), 

                          mean.model = list(armaOrder = c(AR_term, I_term, MA_term)))
    }

}

When I run the above, I get successful messages for many sGARCH models, but eventually get this error: Error: $ operator is invalid for atomic vectors, with the traceback pointing to ugarchfit() and a hessian() function.

I am assuming this is some sort of convergence issue, but have no idea what kind.

EDIT: This is my data (though this same error comes with other datasets as well),

    A
    28.57223993
    28.30616607
    28.2447644
    28.29934366
    28.39485735
    28.80420177
    29.29541506
    29.42504079
    29.31588228
    29.51373208
    30.25737443
    28.94747231
    28.85195861
    28.72915529
    29.17943414
    29.12485489
    29.04298601
    28.96111712
    27.95822332
    28.5381279
    28.68822085
    28.12878349
    27.96504572
    29.32952709
    30.31877609
    30.1345711
    29.629713
    30.01859019
    30.71447569
    30.55756033
    29.09756526
    29.72522669
    29.96401093
    29.96401093
    28.98840675
    27.59663575
    28.07420423
    28.89971546
    28.70868807
    27.75355111
    28.28569885
    29.21354618
    31.89475207
    31.29438027
    31.36260434
    31.41718359

Answer 1

Actually the error appears after very few models. Afterwards many other models throw the same error as well.

It is and isn't a convergence issue. With trace = 1 you can see that in that case hybrid method goes from solnp to nlminb to gosolnp and when, apparently, gosolnp is also unable to get a solution, it fails to exit without errors. The next solver would be nloptr, which actually works fine.

In terms of gosolnp, we have

Trying gosolnp solver...
Calculating Random Initialization Parameters...ok!
Excluding Inequality Violations...
...Excluded 500/500 Random Sequences
Evaluating Objective Function with Random Sampled Parameters...ok!
Sorting and Choosing Best Candidates for starting Solver...ok!
Starting Parameters and Starting Objective Function:
     [,1]
par1   NA
par2   NA
par3   NA
objf   NA

Meaning that all 500 sets of random initial parameters fail to satisfy inequality constraints. As everything else seems to be working fine, I'd suspect that those initial parameter are very unsuitable for GARCH. Trying up to 50000 sets of parameters doesn't help. You could probably experiment with passing distr of gosolnp through solver.control, but that's not great since the same issue arises also with other models (so, likely it's hard to pick a good set of distributions for every case).

So, what we may do is to still use hybrid but to look for an error and if there is one, then to use nloptr:

spec <- spec_creator(x, 'null', MA, AR, I, p, q)
garch <- tryCatch(ugarchfit(spec = spec, data = apple['A'],
                            solver = 'hybrid', solver.control = list(trace = 0)),
                  error = function(e) e)
if(inherits(garch, "error")) {
  garch <- ugarchfit(spec = spec, data = apple['A'],
                     solver = 'nloptr', solver.control = list(trace = 0))
}

I didn't finish running your code with this, but it was fine for over 10 minutes.