Quidest?

Non-parametric One-Way ANOVA

ยท Lorenzo Drumond

1library(coin)
2kruskal_test(Time ~ IDE, data=ide3, distribution="asymptotic") # can't do exact with 3 levels
3kruskal_test(logTime ~ IDE, data=ide3, distribution="asymptotic") # note: same result since based on ranks
4
5#

To follow up on post-hoc test form the omnibus, we execute Mann-Whitneys and then we adjust with Sequential Holm-Bonferroni:

1vs.ec = wilcox.test(ide3[ide3$IDE == "VStudio",]$Time, ide3[ide3$IDE == "Eclipse",]$Time, exact=FALSE)
2vs.py = wilcox.test(ide3[ide3$IDE == "VStudio",]$Time, ide3[ide3$IDE == "PyCharm",]$Time, exact=FALSE)
3ec.py = wilcox.test(ide3[ide3$IDE == "Eclipse",]$Time, ide3[ide3$IDE == "PyCharm",]$Time, exact=FALSE)
4p.adjust(c(vs.ec$p.value, vs.py$p.value, ec.py$p.value), method="holm")

References

#statistics #designing_running_and_analyzing_experiments #non_parametric #assumptions #test #coursera #f_test #experiment #theory #kruskal_wallis #week5 #normality #design #anova #rlang