Table 1.

Key differences between lgpr and LonGP

	lgpr	LonGP (Cheng et al., 2019)
Available kernels	BIN, CAT, ZS, EQ, NS (parameterized warping), VM	BIN, CAT, EQ, PER, NS (fixed warping)
Available observation models	Gaussian, Poisson, NB, binomial, BB	Gaussian
Bayesian inference	Dynamic HMC	Slice sampling and CCD (Vanhatalo et al., 2013)
Heterogeneous effects	Available	Not available
Covariate uncertainty	Available	Not available
Covariate relevance assessment	Decomposition of variance	Stepwise model search with crossvalidation

Note: Kernel name abbreviations: BIN, binary mask; CAT, categorical; ZS, zero-sum; EQ, exponentiated quadratic; NS, non-stationary; VM, variance mask; PER, periodic. The input warping steepness (a in Equation 3) is fixed in LonGP but sampled in lgpr.