The quantitative multiplexing capacity of isobaric Tandem Mass Tags (TMT) has increased the throughput of affinity purification mass spectrometry (AP-MS) to characterize protein interaction networks of immunoprecipitated baits. However, variable bait levels between replicates can convolute interactor identification. We compared the Student's t difference test and Pearson's R correlation as methods to generate t-statistics and assessed the significance of interactors following TMT-AP-MS. Using a simple linear model of protein recovery in immunoprecipitates to simulate reporter ion ratio distributions, we found that correlation-derived t-statistics protect against bait variance while robustly controlling Type I errors (false positives). We experimentally determined the performance of these two approaches for determining t-statistics under two experimental conditions: irreversible prey association to the Hsp40 mutant DNAJB8H31Q followed by stringent washing, and reversible association to 14 3 3 with gentle washing. Correlation-derived t-statistics performed at least as well as difference test t-statistics for each sample, with substantial improvement in performance for experiments with high bait level variance. Deliberately varying bait levels over a large range fails to improve selectivity but does increase robustness between runs. The use of correlation-derived t-statistics should improve identification of interactors using TMT-AP-MS.