... 18

Appendices are only available in electronic form at http://www.aanda.org

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... acceptable

In this context, the H$\beta$ and He II fluxes are poor selection criteria for $\epsilon$, since the former is proportional to the (unknown) covering factor of the shell and the latter depends much on questionable synthetic stellar cluster spectra (Appendix C).

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... low

The tolerance of 9% allowed by SS99 (adopted by V02) is probably too large (Sect. 3.3). This error bar is justified in the logic of SS99, who aim to demonstrate an absence of solution.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... emitted

SS99 prefer casting doubt on the ionization balance of S⁺.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...

Since [O III]${\lambda }$ 5007+4959 is exactly matched, the entry [O III]${\lambda }$ 4363 in Tables 4 and 5 is the "normalized r([O III])'', i.e., the ratio of the computed r([O III]) to the observed r([O III]).

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... for

LP01 state a priori that [O III]${\lambda }$ 4363 "almost surely has a contribution from processes other than photoionization'' and conclude that their model "fails to reproduce the observed [O III]${\lambda }$ 4363 intensity'', but both statements seem to be refuted by evidence they present.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.