Interpreting new Levene Statistic shown by SPSS

I want to perform a between-subjects one way ANOVA on my Data. I have 4 experimental groups. I performed the tests of normality and homogeneity of variances. The output from SPSS for homogeneity of variance is shown below: And the output for Levene's test is shown below: However, the Levene Statistic has 4 rows and I am not sure which one should I consider. Do the test results allow me to perform one-way ANOVA or should I go for Kruskal-Wallis H Test or Welch test?

My total sample size is 14.

I am new to stats. Any help is appreciated.

Thanks.

• Why not simply avoid the assumption of equality of variance in the first place? Then the impact of choosing a test based on a test of assumptions on the properties of your inference is avoided. – Glen_b Jul 20 '18 at 1:45
• @Glen_b Yes, that is also a possibility. If I avoid the assumption of equality of variances then can I represent my results using a between-subjects one way ANOVA in this case? – Jishan Jul 20 '18 at 1:49
• 1. Welch-Satterthwaite ANOVA should be okay in relation to differences in variance; on the other hand, with your design reasonably close to balanced it's probably not much of an impact in any case. 2. You have too few data points to make good within-group assessments of normality; it might be worth discussing what the response variable is and how it's measured as there may be a more suitable analysis than ANOVA – Glen_b Jul 20 '18 at 3:41
• @Glen_b Okay, so basically in my data, I just want to check if there is any statistically significant difference between groups. If yes, then I want to find out which groups differ. My dependent variables are scaled variables – Jishan Jul 20 '18 at 4:50
• 1. Thanks, "longest" is more precise than "different" but we will need further clarification. You'll have a distribution of times, so which will be "longest" may depend on which aspect of the distributions you want to compare across groups. One may be longer by one measure (say mean), a different one longer by a second measure (say median), and a third might be longer by yet another measure (e.g. median pairwise difference across all possible pairs). It could be that each such case is significant at some typical significance level. ... ctd – Glen_b Jul 22 '18 at 1:10