Fuzzy c-medoids clustering on fitted values of splines

cluster.fitted() is used to apply a fuzzy c-medoids clustering algorithm to the fitted values that are output from the TPSfit() function.

Usage

cluster.fitted(
  TPSdata,
  center = TRUE,
  addslopes = TRUE,
  k,
  m = 1.2,
  noise = TRUE,
  seed,
  ...
)

Arguments

TPSdata: Output from the TPSfit() function.
center: A logical expression indicating that trajectories are to be centered on individual means.
addslopes: A logical expression indicating that clustering will be conducted on the slopes between fitted time-points in addition to the cross-sectional values at each point.
k: An integer value indicating the number of clusters.
m: A numeric value called the 'fuzziness parameter.' Must be greater than 1, with a default of 1.2.
noise: A logical expression indicating whether a noise cluster for outliers is to be included.
seed: An optional numeric value for the seed, as the clustering algorithm uses random start values.
...: Additional optional values employed by fclust, including RS, the number of random starts, and delta, the distance for outliers.

Value

An object of class 'FKM.TPS' containing the following components:

FKM_TPS An object output by the fclust algorithm.
U A dataframe containing the degrees of cluster membership for each subject.
Umax A vector containing the modal cluster assignment for each subject.
FKM_TPS_U A data frame containing subjects with subject ID, degrees of cluster membership, and modal class assignment.
FKM_indices A vector containing the values for six cluster validity indices. See fclust::Fclust.index()
wide_data A data frame with the values that were clustered on in the clustering algorithm plus subject IDs.
TPSdata The 'TPSfit' object that was input into the function.
center Logical value that was used to determine whether to center on individual trajectories.
noise Logical value that was used to determine whether a noise cluster is included.
k The number of clusters.
m The fuzziness parameter.
addslopes Logical value that was used to determine whether to include the slopes between fitted time-points.
IDmatch A data frame matching the original subject ID with new consecutive ID values.

Details

cluster.fitted() employs package fclust to apply the fuzzy c-medoids clustering algorithm to the fitted tensor-product spline values that are output from the TPSfit() function. The user has options of including a noise cluster and centering trajectories on the individual means. The addslopes option will cluster on the slopes between fitted time-points in addition to cross-sectional fitting at those points.

Examples

data(TS.sim)

fitsplines <- TPSfit(TS.sim, vars=c("Var1", "Var2", "Var3"), time="Time",
     ID="SubjectID", knots_time=c(0, 91, 182, 273, 365), n_fit_times=10)

clusters1 <- cluster.fitted(fitsplines, k=3, m=1.3, seed=12345, RS=5, noise=TRUE)
summary(clusters1)
#> cluster.fitted(TPSdata = fitsplines, k = 3, m = 1.3, noise = TRUE, 
#>     seed = 12345, RS = 5)
#> 
#> 150 subjects clustered into 3 clusters + noise cluster using m = 1.3 
#> Clusters based on fitted values at times 0 40.44444 80.88889 121.3333 161.7778 202.2222 242.6667 283.1111 323.5556 364 
#> and slopes between points
#> 
#> Cluster summary:
#>       Cl.size Min.degree Max.degree Mean.degree
#> Clus1      47      0.764          1       0.988
#> Clus2      46      0.872          1       0.989
#> Clus3      49      0.698          1       0.990
#> Noise       8      0.957          1       0.990
#> 
#> Component 'FKM_TPS' contains the fuzzy clustering details from package 'fclust'.
#> Components of 'FKM_TPS':
#>  [1] "U"         "H"         "F"         "clus"      "medoid"    "value"    
#>  [7] "criterion" "iter"      "k"         "m"         "ent"       "b"        
#> [13] "vp"        "delta"     "stand"     "Xca"       "X"         "D"        
#> [19] "call"     
#> 
#> Cluster validity indices:
#>         PC         PE        MPC        SIL      SIL.F         XB 
#> 0.92767342 0.03821568 0.89151013 0.35189879 0.37201800 0.22314197 
#> 
#> Output dataset 'FKM_TPS_U' contains degree of cluster membership and modal
#>       cluster assignment for each object.
#> Head of dataset 'FKM_TPS_U':
#>   Id2 SubjectID        Clus1        Clus2        Clus3        Noise ClusModal
#> 1   1        17 5.172103e-07 2.007963e-05 0.9999782952 1.107976e-06         3
#> 2   2        19 9.991481e-01 2.902820e-04 0.0002348860 3.266833e-04         1
#> 3   3        34 7.820025e-08 6.215939e-06 0.9999935927 1.131528e-07         3
#> 4   4        52 2.556221e-07 1.533165e-05 0.9999839290 4.836864e-07         3
#> 5   5        55 9.984985e-01 5.331182e-04 0.0003056236 6.627689e-04         1
#> 6   6        83 9.565796e-01 1.081446e-02 0.0055381485 2.706777e-02         1
#>