K means clustering

Question 1

what is the goal of k means clustering

Answer 1

to find subgroups of participant with maximal within group similarity and minimal between group similarity

Question 2

what factors are is euclidean distance and k means clustering clustering sensitive to

Answer 2

outliers, scale and different random starting centroids

Question 3

what are the two quality criterion in kmeansruns() and what are their goals

Answer 3

ch finds larger more evenly sized clusters. asw forms smaller less evenly sized clusters

Question 4

What is euclidean distance and how is it found

Answer 4

the ordinary straight line distance between two points in a multidimentional space. calculated using pythagorean theorem

Question 5

what metric is used to weigh PCA components in the kmeansCBI() function

Answer 5

square root of the eigenvalue

Question 6

what is the rough process of K means clustering

Answer 6

randomly choose K cases as centroids, assign cases to centroids so sum of euclidean distance is minimised, determine new centroids from clusters, repeat until there is no change or read a predefined maximum

Question 7

why is using PCA components good for k means clustering

Answer 7

uncorrelated and standardised

Question 8

what are two functions used to decide K in K means clustering

Answer 8

nbclust() and kmeansruns()

Question 9

what function is used to find which clusters stand up to resampling

Answer 9

kmeansCBI()

Question 10

to what part of the clusterboot() function is the output of kmeansCBI added to

Answer 10

scaling =

Question 11

what does a jaccard similarity show

Answer 11

how well each cluster from the best cluster solution replicated over the number of bootstrap samples

Question 12

what is a reasonably robust jaccard similarity

Answer 12

0.6

Question 13

what are two factors to be balanced when deciding K, and what metric is used to consider them

Answer 13

robustness of clusters (jaccard similarity) and variability explained (between_ss/total_ss)