Testing for SNPs predictive of clinical outcome often starts with testing for different genetic effects (such as dominant, recessive, and additive) within each SNP. Then the minimum p-value of these tests is taken to represent the potential significance of that SNP, and all the SNPs are then ranked accordingly. Concentrating on the starting point of this practice, testing within each SNP, we suggest an alter-native approach that is more informative for the purpose of drug development. For a single SNP, we provide simultaneous confidence intervals for dominant, recessive, and additive effects on the clinical response. In Type II diabetes, for instance, they would be confidence intervals for mean difference between treatment and control of glycosylated hemoglobin (HbA1c) reduction from baseline for all tested effects. This simultaneous confidence intervals method is more informative for the following reason. A reduction in HbA1c between 0.8 and 1.2 is much more clinically meaningful than a reduction between 0.4 and 0.6. Yet the confidence intervals (0.8, 1.2) and (0.4, 0.6) can have identical p-values. The improvement of the within SNP inference is crucial for a better overall inference across the SNPs in assessing treatment efficacy.