i’m a biologist, and i’m interested in morphological differences between individuals

Currently, i’m working with something called aspect ratio of the caudal fin. I want to identify if individuals with similar size have different aspect ratios.

However, the aspect ratio is long known to have a positive correlation with fish size. Therefore, my first step was remove the influence of size in aspect ratio by creating a new parameter: the resid from a simple linear model between aspect ratio and fish size.

Then, i wanted to identify different groups in my datasets based on this resid values. For that I created a subset with only the resid parameter, and i thought clustering would be the best way to categorize individuals in groups.

My datasets:

           | Dataset |  nrows  |
           | ------- | ------- |
           |  1      |   101   |
           |  2      |   55    |
           |  3      |   102   |

Knowing my datasets (each one a species) are small, to determine the optimal number of clusters (k) i’ve applied the NbClust function with a hierarchical method.

NbClust(subset, min.nc = 2, max.nc = 10, method = "ward.D2")

Once with the num_clusters, I did the analysis, plotted the dendrogram, eliminated the outliers and repeated the process.

dist_matrix <- dist(subset)
hc <- hclust(dist_matrix, method = "ward.D2")

plot(hc, main = paste("Dendrograma com", num_clusters, "Clusters"),
     xlab = "Observações", ylab = "Distância Euclidiana")

rect.hclust(hc, k = num_clusters, border = "red")

Then, i transformed the clusters in categories in my original dataset

clusters <- cutree(hc, k = num_clusters)
hast$Cluster <- clusters
hast$Cluster<-factor(hast$Cluster)

Finally, i wanted to compare the clusters of individuals

anova<-aov(resid~Cluster, data = dataset)
TukeyHSD(anova)

I don’t know if my approach is correct, neither if clustering is the better option. Therefore, i want help from people more experienced in statistics