The well-known microactivity test (MAT) ASTM D 3907-92, has been used worldwide in laboratory scale testing to determine the activity and selectivity of fluid catalytic cracking (FCC) catalysts. To assess catalytic selectivities by interpolation, either a kinetic model or functional (empirical) relationships between product yields and conversion should be used. Experimental errors may significantly affect analysis and erroneous conclusions might arise, justifying application of statistical methods. Although these methods are well established, they are frequently neglected or oversimplified in catalysis. The aim of the present work is to evaluate the influence of typical experimental errors from MAT units on interpolated yields, based on previously tested empirical functions and data from the literature, and from FCC catalyst samples from PETROBRAS Research Center, with and without a vanadium trap additive. Statistical criteria to identify possible outliers and the confidence intervals for predicted values are discussed. It is shown that the experimental errors on the independent variable, sometimes disregarded, drastically decrease the discrimination capacity of the MAT test concerning iso-conversion and iso-coke evaluations. Concentrating experiments near the conversion region of interest and improvement of MAT repeatability could mitigate this unavoidable problem. These conclusions can be generalized to selectivity analyses related to the reaction systems other than catalytic cracking.