An alternative to the conventional ANOVA that does not require equal variances is called Welch’s ANOVA. When there is variable variance, it may be more resilient.

When the premise of equal variances is broken, Welch’s ANOVA is made to be resilient. By adjusting the conventional F-statistic to take unequal variances into account, it achieves this robustness. The ratio of the mean squares is used to compute the Welch F-statistic, which accounts for various group variances.

In my analysis earlier during ANOVA, the Levene’s test revealed significant differences in variances between ethnic groups, casting doubt on the fundamental premise of equal variances in conventional ANOVA. I chose Welch’s ANOVA as a substitute because I understood how its unique design addressed different variances by modifying the F-statistic and degrees of freedom for greater precision. However, the Shapiro-Wilk tests indicated that the age distribution was not normal in the majority of racial groups. I recognise that large deviations from normalcy may affect results, even though Welch’s ANOVA can withstand mild deviations. The degree of these violations and the particulars of my dataset will determine whether I use Welch’s ANOVA or look into other options. I might think about using non-parametric tests in situations when there is extreme non-normality or small sample numbers.