Closed zh-zhang1984 closed 2 years ago
Thanks for your question. The documentation of the lcMethodLMKM
function is indeed lacking, I will extend it.
lcMethodLMKM
documentationIn the meantime, I'll try to answer your question:
lcMethodLMKM
is an implementation of a type of individual time series (ITS) approach, also known as a feature-based approach, where each trajectory is represented by the coefficients of an individually fitted linear regression model. The trajectories are then clustered based on the coefficients using k-means. LMKM is not an established name or specific method, so I don't have any references for it.
I've described the feature-based approach, although not LMKM specifically, in more detail in a general tutorial on longitudinal clustering: https://arxiv.org/abs/2111.05469, see Section 4.3
A linear regression representation approach is also used in Anchored k-medoids (see lcMethodAkmedoids), but as far as I'm aware the package does not enable you to specify the linear regression model to use.
The LMKM method is similar to GCKM (as implemented in lcMethodGCKM
), except that for GCKM the coefficients are determined using a linear mixed model (a growth curve model), which should be more robust.
GCKM has been described and compared in:
thank you for the excellent explanation and references. I will read into it.
Dear authors, I am confusing on the methodology behind lcMethodLMKM function, are there any hints /reference for this procedure? I searched in internet but find nothing to be relevant.