We aim to investigate the propriety of stellar parameter errors of the official data release of the LAMOST low-resolution spectroscopy (LRS) survey. We diagnose the errors of radial velocity (RV), atmospheric parameters ([Fe/H], T log g) and α-enhancement ([α/M]) for the latest data release version of DR7, including 6,079,235 effective spectra of 4,546,803 stars. Based on the duplicate observational sample and comparing the deviation of multiple measurements to their given errors, we find that, in general, the error of [α/M] is largely underestimated, and the error of RV is slightly overestimated. We define a correction factor k to quantify these misestimations and correct the errors to be expressed as proper internal uncertainties. Using this self-calibration technique, we find that the k-factors significantly vary with the stellar spectral types and the spectral signal-to-noise ratio (S/N). Particularly, we reveal a strange but evident trend between k-factors and error themselves for all five stellar parameters. Larger errors tend to have smaller k-factor values, i.e., they were more overestimated. After the correction, we recreate and quantify the tight correlations between S/N and errors, for all five parameters, while these correlations have dependence on spectral types. It also suggests that the parameter errors from each spectrum should be corrected individually. Finally, we provide the error correction factors of each derived parameter of each spectrum for the entire LAMOST-LRS DR7 and plan to update them for the later data releases.
It accepts original submissions from all over the world and is internationally published and distributed by IOP