posterior_epred fails for stan_clogit model with rms::rcs splines when newdata has only one value for spline covariate

[epiben]

Summary:

After fitting a stan_clogit model in which two covariates are modelled with restricted cubic splines, posterior_epred fails when trying to find conditional effects of one of these covariates if they other is fixed at one value.

Description:

I'm fitting a conditional logistic regression to a data set in which each stratum has 1 case and 5 controls, mathed on date of birth +/- 180 days. I use restricted cubic splines for the most interesting continuous covariates; a few categorical covariates are encoded as factors in the data input. When using the posterior_epred function to get conditional effects I've encountered two problems. I've opened an issue for one here; this issue addresses other one: in a model with two covariates modelled with restricted cubic splines (specifically, rms::rcs), posterior_epred fails when giving newdata in which one of the spline covariates has a range of values (over which to evaluate the conditional effect) and the other is fixed at one value.

The problem seems to be that posterior_epred tries to create splines based on the newdata input, but this is of course not meaningful with a single-valued covariate.

Reprex:

Prep work:

library(rstanarm)
library(dplyr)

dat <- arrange(infert, stratum) %>% 
	filter(parity <= 2) %>% 
	mutate(test_covar = rnorm(n(), 10, 3))

nd <- data.frame(age = 18:45, spontaneous = 0, induced = 0, stratum = "1", 
                 case = c(1, rep(0, 27)), test_covar = 8)

Works:

post <- stan_clogit(case ~ rms::rcs(age) + test_covar + spontaneous + induced, 
                    strata = stratum, data = dat, QR = TRUE)
mu <- posterior_epred(post, newdata = nd) 
rowSums(mu) # all 1

Fails (the model trains):

post2 <- stan_clogit(case ~ rms::rcs(age) + rms::rcs(test_covar) + spontaneous + induced, 
                     strata = stratum, data = dat, QR = TRUE)
mu2 <- posterior_epred(post2, newdata = nd)

with this message:

fewer than 3 unique knots.  Frequency table of variable:
x
 8 
28 
Error in rcspline.eval(x, nk = nknots, inclx = TRUE, pc = pc, fractied = fractied) : 
  
In addition: Warning message:
In rcspline.eval(x, nk = nknots, inclx = TRUE, pc = pc, fractied = fractied) :
  5 knots requested with 1 unique values of x.  knots set to -1 interior values.

Versions and OS:

I run rstanarm v2.21.1 in Rstudio with R v3.6.3 on MacOS X 10.15 Catalina.