difference robust = true/false

[philippcla]

Hello

I have a dataset at a certain frequency. The resulting plot of

gmwm(gmwm_model, imu_object, freq = 200, robust = FALSE,  seed = 1)

is different from this

gmwm(gmwm_model, imu_object, freq = 200, robust = TRUE,  seed = 1)

it differs in the "empirical WV", why is that, due to the robust flag, should it change so dramatically? The last scale is really "off"

And because of this, the parameters are different too.

robust = true

Model Information: 
          Estimates
AR1    9.752502e-01
SIGMA2 1.382484e-08
AR1    2.680622e-01
SIGMA2 4.016523e-04
WN     5.681748e-05
QN     2.331227e-06

robust = false

Model Information: 
          Estimates
AR1    9.999821e-01
SIGMA2 2.189146e-11
AR1    2.705570e-01
SIGMA2 4.093538e-04
WN     4.308224e-05
QN     5.355913e-06

[coatless]

WV Empirical should be different between robust & classical types if the efficiency supplied is not equivalent to 1, in which case robust estimate is the same as the classical.

Try:

wv.c = wvar(imu_object) # classical
wv.r = wvar(imu_object, robust = T) # robust
compare_wvar(wv.c, wv.r)

I'm thinking that the issue for the parameter estimates is volatility in the latter scales.

Mind sending over the data?

[philippcla]

Hello,
I used the following things:

gmwm_model <- 2*AR1()+WN()+QN()

imu_object_hot_TRUE <- imu(data = hot_data, acc = 1, freq = 200)
estimated_model_hot_TRUE <- gmwm(gmwm_model, imu_object_hot_TRUE, freq = 200, robust = TRUE,  seed = 1)
plot(estimated_model_hot_TRUE,  process.decomp = TRUE, title = "TRUE"))
summary(estimated_model_hot_TRUE)

imu_object_hot_FALSE <- imu(data = hot_data, acc = 1, freq = 200)
estimated_model_hot_FALSE <- gmwm(gmwm_model, imu_object_hot_FALSE, freq = 200, robust = FALSE,  seed = 1)
plot(estimated_model_hot_FALSE,  process.decomp = TRUE, title = "FALSE"))
summary(estimated_model_hot_FALSE)

and here is the data in a zip-file:
hot_data_axis6.RData.zip

thx