Turn sdmTMB model output into a tidy data frame
Arguments
- x
Output from
sdmTMB().- effects
A character value. One of
"fixed"('fixed' or main-effect parameters),"ran_pars"(standard deviations, spatial range, and other random effect and dispersion-related terms),"ran_vals"(individual random intercepts or slopes and time-varying coefficients, if included; behaves likeranef(); see Details),"ran_vcov"(list of variance covariance matrices for the random effects, by model and group), or"rsr"(Restricted Spatial Regression fixed-effect coefficients adjusted for spatial confounding with the random fields; Hanks et al. 2015; Diaz and Thorson 2025). To access RSR coefficients the model must be fitted withcontrol = sdmTMBcontrol(get_rsr = TRUE)."dispersion"returns dispersion-model coefficients whendispformulais used.- model
Which model to tidy if a delta model (1 or 2). The
modelwill be ignored when effects is"ran_vals"(all returned in a single data frame with amodelcolumn).- conf.int
Include a confidence interval?
- conf.level
Confidence level for CI.
- exponentiate
Whether to exponentiate the fixed-effect coefficient estimates and confidence intervals.
- silent
Omit any messages?
- ...
Extra arguments (not used).
Details
Follows the conventions of the broom and broom.mixed packages.
Currently, effects = "ran_pars" also includes dispersion-related terms
(e.g., phi) only when dispersion is scalar. With dispformula,
use effects = "dispersion" to extract dispersion-model coefficients.
Time-varying coefficients (from the time_varying argument in sdmTMB())
are returned with effects = "ran_vals", one row per coefficient and time
slice, with terms named "<coefficient>:<time>" (e.g., "(Intercept):2011").
These are on the link scale. Their standard deviation (sigma_V) and, for
time_varying_type = "ar1", correlation (rho_time) are returned with
effects = "ran_pars".
For models with penalized smoothers, effects = "ran_pars" includes an edf
column with the effective degrees of freedom of each smoother (NA for
other parameters). As in mgcv, this counts the unpenalized null space
of the smoother (e.g., 1 for the linear component of s(x)) plus the
effective number of penalized coefficients, integrating over the fixed and
random effects and conditioning on the variance parameters. With t2(), the
smoother's null space is added to its first penalized component so the
components sum to the mgcv term EDF. These are computed when the
model is fit, typically at a small fraction of the fitting time. They differ from
cAIC(fit, what = "EDF"), which conditions on the fixed effects and excludes
the null space.
With anisotropy = TRUE, effects = "ran_pars" reports range_min and
range_max, the ranges along the minor and major axes, in place of range.
Their geometric mean is range. Their standard
errors are from the delta method and their confidence intervals are Wald
intervals on the log scale. These intervals are unreliable when the
anisotropy is weakly identified, such as when range_max / range_min is
close to 1, because the axis ranges have no derivative at isotropy. They are
NA at exact isotropy.
Standard errors for spatial variance terms fit in log space (e.g., variance terms, range, or parameters associated with the observation error) are omitted to avoid confusion. Confidence intervals are still available.
References
Restricted Spatial Regression (effects = "rsr"):
Diaz, R.R., and Thorson, J.T. 2025. When and How to Use Restricted Spatial Regression to Separate Environmental Effects from Spatial Confounding. EcoEvoRxiv. doi:10.32942/X28351 .
Hanks, E.M., Schliep, E.M., Hooten, M.B., and Hoeting, J.A. 2015. Restricted spatial regression in practice: geostatistical models, confounding, and robustness under model misspecification. Environmetrics 26(4): 243–254. doi:10.1002/env.2331 .
Examples
fit <- sdmTMB(density ~ poly(depth_scaled, 2, raw = TRUE),
data = pcod_2011, mesh = pcod_mesh_2011,
family = tweedie()
)
tidy(fit)
#> # A tibble: 3 × 5
#> term estimate std.error conf.low conf.high
#> <chr> <dbl> <dbl> <dbl> <dbl>
#> 1 (Intercept) 3.65 0.281 3.10 4.20
#> 2 poly(depth_scaled, 2, raw = TRUE)1 -1.54 0.186 -1.90 -1.17
#> 3 poly(depth_scaled, 2, raw = TRUE)2 -1.11 0.101 -1.31 -0.913
tidy(fit, conf.int = TRUE)
#> # A tibble: 3 × 5
#> term estimate std.error conf.low conf.high
#> <chr> <dbl> <dbl> <dbl> <dbl>
#> 1 (Intercept) 3.65 0.281 3.10 4.20
#> 2 poly(depth_scaled, 2, raw = TRUE)1 -1.54 0.186 -1.90 -1.17
#> 3 poly(depth_scaled, 2, raw = TRUE)2 -1.11 0.101 -1.31 -0.913
tidy(fit, "ran_pars", conf.int = TRUE)
#> # A tibble: 4 × 5
#> term estimate std.error conf.low conf.high
#> <chr> <dbl> <dbl> <dbl> <dbl>
#> 1 range 19.1 14.0 4.58 80.0
#> 2 phi 14.0 0.677 12.8 15.4
#> 3 sigma_O 2.14 0.941 0.906 5.07
#> 4 tweedie_p 1.58 0.0153 1.55 1.61
pcod_2011$fyear <- as.factor(pcod_2011$year)
fit <- sdmTMB(density ~ poly(depth_scaled, 2, raw = TRUE) + (1 | fyear),
data = pcod_2011, mesh = pcod_mesh_2011,
family = tweedie()
)
tidy(fit, "ran_vals")
#> # A tibble: 4 × 7
#> group_name term level_ids estimate std.error conf.low conf.high
#> <chr> <chr> <chr> <dbl> <dbl> <dbl> <dbl>
#> 1 fyear (Intercept) 2011 0.0163 0.187 -0.351 0.384
#> 2 fyear (Intercept) 2013 0.177 0.188 -0.192 0.545
#> 3 fyear (Intercept) 2015 0.232 0.189 -0.139 0.603
#> 4 fyear (Intercept) 2017 -0.431 0.205 -0.833 -0.0287
# time-varying coefficients:
fit <- sdmTMB(density ~ 0, time_varying = ~ 1, time = "year",
data = pcod_2011, mesh = pcod_mesh_2011, family = tweedie(),
spatiotemporal = "off"
)
tidy(fit, "ran_vals", conf.int = TRUE)
#> # A tibble: 4 × 5
#> term estimate std.error conf.low conf.high
#> <chr> <dbl> <dbl> <dbl> <dbl>
#> 1 (Intercept):2011 2.94 0.495 1.97 3.91
#> 2 (Intercept):2013 3.07 0.490 2.11 4.03
#> 3 (Intercept):2015 3.16 0.495 2.19 4.13
#> 4 (Intercept):2017 2.66 0.501 1.68 3.64
tidy(fit, "ran_pars")
#> # A tibble: 5 × 5
#> term estimate std.error conf.low conf.high
#> <chr> <dbl> <dbl> <dbl> <dbl>
#> 1 range 38.2 17.0 16.0 91.3
#> 2 phi 15.4 0.719 14.1 16.9
#> 3 sigma_O 2.02 0.319 1.48 2.75
#> 4 sigma_V 0.348 0.193 0.117 1.03
#> 5 tweedie_p 1.59 0.0154 1.56 1.62
