including p-values in customized output omits random effects from mixed models

[strengejacke]

model <- lme4::lmer(Sepal.Width ~ Petal.Length + (1|Species), data = iris)
modelsummary::modelsummary(model, estimate  = "{estimate} [{conf.low}, {conf.high}] {p.value}", statistic = NULL, fmt = 2)

Current output:

It would be great to keep the random effects, and just set blank fields for p-values in the random effects rows.

[vincentarelbundock]

Thanks for reporting! I don’t use mixed-effects models often, so these are very useful test cases.

Should be fixed in 0959b02

Note that user-input in glue strings is totally arbitrary, so my current checks won’t catch every possibility, but it should be a considerable improvement.

library(lme4)
library(modelsummary)
model <- lmer(Sepal.Width ~ Petal.Length + (1|Species), data = iris)
modelsummary(model,
    "markdown",
    estimate  = "{estimate} [{conf.low}, {conf.high}] {p.value}",
    statistic = NULL)

	Model 1
(Intercept)	1.998 [0.888, 3.109] 0.001
Petal.Length	0.282 [0.165, 0.399] 0.000
SD (Intercept Species)	0.892
SD (Observations)	0.316
Num.Obs.	150
RMSE	0.31

[strengejacke]

Nice, thanks!

[strengejacke]

I'm re-opening it, because it doesn't fully work when CIs are only partially available, see:

m1 <- lme4::lmer(Sepal.Width ~ Petal.Length + (1|Species), data = iris)
m2 <- lme4::lmer(Sepal.Width ~ Petal.Length + (1 + Petal.Length |Species), data = iris)
m3 <- lme4::lmer(Sepal.Width ~ Petal.Length + Petal.Width + (1 + Petal.Length |Species), data = iris)
#> boundary (singular) fit: see help('isSingular')
parameters::compare_models(m1, m2, m3, effects = "all")
#> Your model may suffer from singularity (see '?lme4::isSingular' and
#>   '?performance::check_singularity').
#>   Some of the standard errors and confidence intervals of the random
#>   effects parameters are probably not meaningful!
#> Parameter                    |                m1 |                     m2 |                    m3
#> -------------------------------------------------------------------------------------------------
#> (Intercept)                  | 2.00 (0.89, 3.11) | 1.99 ( 0.89,     3.10) | 1.76 (0.44,     3.08)
#> Petal Length                 | 0.28 (0.16, 0.40) | 0.29 ( 0.16,     0.41) | 0.16 (0.03,     0.29)
#> SD (Intercept)               | 0.89 (0.33, 2.43) | 0.88 ( 0.27,     2.87) | 1.10 (0.36,     3.30)
#> SD (Observations)            | 0.32 (0.28, 0.35) | 0.32 ( 0.28,     0.35) | 0.30 (0.26,     0.33)
#> SD (Petal.Length)            |                   | 0.02 ( 0.00, 1.76e+07) | 0.03 (0.00, 2.33e+05)
#> Cor (Intercept~Petal.Length) |                   | 0.19 (-1.00,     1.00) | 1.00 (    ,     1.00)
#> Petal Width                  |                   |                        | 0.61 (0.34,     0.88)
#> -------------------------------------------------------------------------------------------------
#> Observations                 |               150 |                    150 |                   150

^{Created on 2022-06-08 by the reprex package (v2.0.1)}

In this case, the following code produces following table:

modelsummary::modelsummary(model, estimate  = "{estimate} [{conf.low}, {conf.high}]", 
                           shape = group + term + statistic ~ model,
                           statistic = NULL, fmt = 2)

Note the message from compare_parameters() (requires this PR to work, I think).

[vincentarelbundock]

Is this working as you expected with the latest commits?

library(modelsummary)
models <- list(
    lme4::lmer(Sepal.Width ~ Petal.Length + (1|Species), data = iris),
    lme4::lmer(Sepal.Width ~ Petal.Length + (1 + Petal.Length |Species), data = iris),
    lme4::lmer(Sepal.Width ~ Petal.Length + Petal.Width + (1 + Petal.Length |Species), data = iris))
modelsummary(
    models,
    estimate = "{estimate} [{conf.low}, {conf.high}]",
    statistic = NULL,
    gof_map = NA,
    output = "markdown")

	Model 1	Model 2	Model 3
(Intercept)	1.998 [0.888, 3.109]	1.993 [0.890, 3.096]	1.762 [0.439, 3.084]
Petal.Length	0.282 [0.165, 0.399]	0.286 [0.164, 0.408]	0.160 [0.027, 0.292]
Petal.Width			0.610 [0.341, 0.878]
SD (Intercept Species)	0.892 [0.328, 2.430]	0.885 [0.273, 2.866]	1.096 [0.364, 3.300]
SD (Petal.Length Species)		0.024 [0.000, 17569330.944]	0.026 [0.000, 232874.419]
Cor (Intercept~Petal.Length Species)		0.189 [-1.000, 1.000]	1.000 [, 1.000]
SD (Observations)	0.316 [0.281, 0.354]	0.315 [0.281, 0.354]	0.296 [0.264, 0.332]

^{Created on 2022-06-08 by the reprex package (v2.0.1)}

[vincentarelbundock]

Preparing a Ripley-induced emergency release, so closing all the issues which seem (to me) complete. Feel free to re-open or keep the discussion going if you see a need.