First of all, I would like to apologise if this question has already been answered multiple times. I do realise plenty of similar questions have been answered here, but I feel that these do not sufficiently answer my question.

I have a dataset with numerous measurements from 76 individuals, originating from Belgium, Germany, and Switzerland, of which 27 are female, 49 are male. I want to test the effects of sex and origin, as well as the interaction between these factors, on the outcomes in R. From Belgium and Switzerland I have both males and females, but from Germany I have only males.

I have looked at the effects of the various ANOVA types, but they all present the exact same outcome. If I remove all German individuals, and run an ANOVA on the other two countries, I do get different results. So I do understand that R is not excluding my Germans pairwise.

results<-data.frame()
for (i in 5:20){
   aov_mid<-aov(sample[,i]~Sex*Site,data=sample)
   aov_end<-Anova(aov_mid,Type="III")
   results[c(1:3),i-4]<-aov_end[c(1:3),4]
   tukey<-TukeyHSD(aov_mid)
   if (aov_end[3,4]<0.05){
      results[c(5:10),i-4]<-tukey$'Sex:Site'[c(1:6),4]
      }
colnames(results)[i-4]<-colnames(sample)[i]
}

Some studies advise the use of ANOVA Type IV in such cases, while others suggest the use of mixed models. However, most examples deal with longitudinal data or repeated measures, which is not really applicable here.

Alternatively, I could assess the effects of origin by running a t-test on the Swiss and Belgian females, and a separate one-way ANOVA on the males, but this would still leave the question of the interaction unanswered. Similarly, I have tried leaving out the Germans to test the interaction, but the outcome will be much more restricted.

Thanks in advance for any help and advice offered!

0 Answers