All of the data reductions were performed using
IRAF
.
The IRAF packages
CCDRED, TWODSPEC and ONEDSPEC were used to reduce the long-slit spectral
data. The CCD reductions included bias subtraction, flat-field correction and
cosmic-ray removal. The dark counts were so low that their subtraction was not performed.
We adopted Horne's (1986) avariance weighed algorithm in order to extract the spectra.
As for the aperture size, we adopted similar method to that of
Kim et al. (1995) to minimize the aperture-related effects on the nuclear spectra. The
apertures were varied according to the redshift of the objects so that they approached
diameters of 2 kpc 
for galaxies with z < 0.034 , but were fixed
at 4'' for objects at larger redshifts as the seeing was typically
3'' to 4'' and the pixel
sizes was 1.3'' . The only two exceptions were for
and 
(M 82) which
are quite nearby and a 2 kpc aperture was too large for them.
An Fe-He-Ar lamp was used for the wavelength calibrations.
More than 20 lines were used to establish the wavelength scale by fitting a first-order spline3
function. The accuracy of the wavelength calibration
was better than 
.
On most nights, more than two KPNO standard stars were used to perform the relative flux calibration. Atmospheric extinction was corrected using the mean extinction coefficients for the Xinglong station, that were measured out by BATC multi-color survey (Yan 1995).
The telluric 
absorption bands at 
and 
did heaver effect on those emission lines at the similar observed wavelength.
We therefore constructed an artificial - the spectrum
of standard stars setting every wavelength to unity except these wavelength corresponding to the bands.
Division by the artificial spectrum could therefore removed telluric absorption bands.
In most cases, this technique worked well,
especially for the band near . However in
some cases, this correction seems not to have been very satisfactory and affected the measurement
of the emission lines at .