<p>The traditional analysis of variance (ANOVA) in artificial intelligence (AI) is used to test whether group means differ significantly. However, classical ANOVA cannot handle environments where uncertainty exists. In this paper, we introduce a modified version of ANOVA, called neutrosophic analysis of variance (NANOVA), which is applicable in AI when uncertainty is present. NANOVA addresses cases where training and testing data contain imprecise observations. We conducted a detailed simulation study to examine how varying degrees of indeterminacy affect the descriptive statistics and <i>F</i>-values for both training and testing data. The results from the proposed NANOVA are compared with classical ANOVA to demonstrate the impact of uncertainty on analysis outcomes. An example application of NANOVA using AI training and testing data is provided, showcasing its practical relevance. Our findings reveal that the degree of indeterminacy significantly alters the results, affecting the means, variances, and <i>F</i>-values. This highlights the necessity of considering uncertainty in statistical analysis for AI, as traditional methods may not fully capture the complexities of uncertain environments. By incorporating NANOVA, AI researchers can better account for indeterminate data, leading to more accurate and reliable decision-making in uncertain scenarios.</p>