Figure 4

Comparison of the output generated by the different methods considered for clustering the clinical variable HOMA IR time profiles. The methods were set for n = 2 clusters, using data for both genders and age range 11–16. (a,b) Mfuzz and kmInMfuzz respectively for the fuzzy logic and the k-means algorithm (Mfuzz package); (c) SSClust for the Smoothing Spline Clustering method (SSC); (d) longclust for the model based clustering for longitudinal data; (e,f) nbclust1 and nbclust3 respectively for the clustering schemes described in the text; (g,h) tscluCORThc and tscluCORTpam respectively for the clustering schemes described in the text.