We show that in order to minimize the uncertainties in the N and O abundances of low mass , low metallicity ( O/H \leq solar/5 ) emission-line galaxies , it is necessary to employ separate parameterizations for inferring T _ { e } ( N ^ { + } ) and T _ { e } ( O ^ { + } ) from T _ { e } ( O ^ { +2 } ) .
In addition , we show that for the above systems , the ionization correction factor ( ICF ) for obtaining N/O from N ^ { + } /O ^ { + } , where the latter is derived from optical emission-line flux ratios , is \left < ICF \right > = 1.08 \pm 0.09 .
These findings are based on state-of-the-art single-star H II region simulations , employing our own modeled stellar spectra as input .
Our models offer the advantage of having matching stellar and nebular abundances .
In addition , they have O/H as low as solar/50 ( lower than any past work ) , as well as log ( N/O ) and log ( C/O ) fixed at characteristic values of -1.46 and -0.7 , respectively .
The above results were used to re-derive N and O abundances for a sample of 68 systems with 12 + log ( O/H ) \leq 8.1 , whose de-reddened emission-line strengths were collected from the literature .
The analysis of the log ( N/O ) versus 12 + log ( O/H ) diagram of the above systems shows the following : ( 1 ) the largest group of objects forms the well-known N/O plateau with a value for the mean ( and its statistical error ) of -1.43 ( +.0084/-.0085 ) ; ( 2 ) the objects are distributed within a range in log ( N/O ) of -1.54 to -1.27 in Gaussian fashion around the mean with a standard deviation of \sigma = +.071/-.084 ; and ( 3 ) a \chi –square analysis suggests that only a small amount of the observed scatter in log ( N/O ) is intrinsic .