I have been working on the validation of a raw material method. The recovery was performed and found the assay values would decrease in as the concentration was increased (97.4 at 80%, 96.3 at 100% and 94.8 at 120%). This appeared to be at the max of the linearity curve. The linearity was performed (40, 80, 100, 120, and 160%) around the sample concentration and gave a straight line (0.9998 corr.).

Typically the recovery would show if there are any matrix affects of a dose drug. But, this is a raw material.

Any ideas would be greatly appreciated. Thanks in advance.

Standard deviation of response factors gives a much better indication of linear than correlation coefficient. Response factors give the slope of the line drawn from the origin to each data point. The response factors will be the same only if the response as a function of concentration is linear, and the blank value is 0. Correlation measures how well points follow a trend, and is not necessarily a measure of true linearity. Ideally the response factor standard deviation will be small and the correlation close to 1. It is impossible to have a low correlation coefficient if the standard deviation of response factors is a low value, but it is possible to have a good correlation and SD of response factors of 30% or more.

To sum up, a curve the correlation coefficient indicates is linear may not truly be linear, and this may be the cause of some of the apparent conflict you see.

Thank you very much for stating what I should have known. I put the data from the recovery into a linearity calculation and also had a correlation of 0999.

Would you say that the sample is not behaving according to Beer's law at this concentration?

If this is a chromatographic method perhaps the absorbance of the analyte at the wavelength you have chosen is not in the linear range of the detector. As a rule of thumb I generally keep all absorbances at 1 or less during quantitative analysis.