test_that("Prediction works with missing time", { skip_on_cran() fit <- sdmTMB( density ~ 1, data = pcod_2011, mesh = pcod_mesh_2011, time = "year", family = tweedie(link = "log") ) nd <- pcod_2011[pcod_2011$year %in% c(2013, 2017), ] p1 <- predict(fit, newdata = nd) p2 <- predict(fit, newdata = pcod_2011) p2 <- p2[p2$year %in% c(2013, 2017), ] expect_equal(nrow(p1), nrow(p2)) expect_equal(p1$est, p2$est) expect_equal(p1$year, p2$year) expect_equal(p1, p2) })