MarkovRegime Heteroskedasticity; implied st dev Eviews Help

[E K]

Hi there,
I am trying to solve, how the implied standard deviation was calculated in the Eviews Help file for Markov Regime Switching
http://www.eviews.com/EViews8/ev8ecswitch_n.html#RegHet

How did they retrieve the values for implied stand dev. 0.035, 0.176, and 0.063, respectively?
Here is what the help files says:

"Kim and Nelson (1999) offer an example (Section 4.6, p. 86) of a three state Markov switching model of regime heteroskedastic stock returns from 1926m1–1986m12. The data, which consist of monthly CRSP equal-weighted excess returns are in the series EXCESS, provided in the workfile ew_excs.WF1.

The specification of the model is as depicted in the dialog below:
Switching Regression Regime Heteroskedasticity

The excess returns are assumed to have mean zero so we enter only the name of the dependent variable in the topmost edit field. Since we wish to model regime heteroskedasticity, Regime specific error variances box is checked. The model assumes Markov switching probabilities with 3 regimes and constant transition probabilities.

Preliminary analysis indicates that this model is particularly difficult to estimate with a number of local roots exhibiting coefficient singularity.

To obtain estimates we instruct EViews to perform extra randomized starting value estimation. Click on the Options tab and change the starting value settings so EViews generates 50 sets of random starts (instead of 25) with 20 (instead of 10) iteration refinements. Click on OK to proceed with the estimation.

EViews estimates the model and displays the standard switching regression output:
Switching Regression

The results show estimates of the log standard deviations in the low, high, and medium volatility regimes. The implied standard deviations are 0.035, 0.176, and 0.063, respectively.

Thank you all in advance!

[EViews Glenn]

The results show the coefficient estimates for the log of sigma for each of the three regimes. Taking the exponential of the three values gives the three values for sigma.

[E K]

Doh! I should have explored the underlying dataset more thoroughly.
Thank you very much!
This was very helpful.
THX!