Conventionally, Immunoassay limit of detection (LoD) and limit of quantitation (LoQ) were evaluated based on distribution of measured concentration. The LoD and LoQ were low and look like more sensitive, but often lower than linear range. In CLSI EP17-A2, there is another modeling/regression option. There is a concentration mean polynomial model, Std = B0` + B1`*Mean + B2'*Mean**2, for LoD and a concentration mean power function model, %CV = B0*Mean**B1, for LoQ, where LoQ is general within linear range. A model relative standard deviation (%RSD) or model CV is 12.5%. AIC=-95.86. If the LoD equation is divided by Mean at both sides, Std/Mean = B0'/Mean + B1' +B2'*Mean is deduced. From this equation, we proposed mean inverse model approach for LoQ, %CV=B0+B1/Mean+B2*Mean or %CV=B0+B1/Mean, if B2 p-value is statistical insignificant. Allowable %CV=10% and previous equation to determine LoQ. The %RSD of mean inverse model is 5.7% using same data of power function model and AIC=-96.28. While Total error is TotErr=Bias+2Std, Bias+2Std =B0` + B1`*Mean. (Bias+2Std)/Mean= B0'/Mean + B1', if divided by Mean both sides. PctTE=PctBias+2CV. PctBias + 2CV = 5% + 2*10%. Allowable PctTEa=25% and PctTE = B0 + B1/Mean to determine LoQ. The %RSD of mean inverse model is much smaller and more mathematical meaningful than the power function model.