Marginal Contrasts with marginaleffects::avg_comparisons()
Source: R/marginal_tidiers.R
tidy_marginal_contrasts.RdUse marginaleffects::avg_comparisons() to estimate marginal contrasts for
each variable of a model and return a tibble tidied in a way that it could
be used by broom.helpers functions.
See marginaleffects::avg_comparisons() for a list of supported models.
Usage
tidy_marginal_contrasts(
x,
variables_list = "auto",
conf.int = TRUE,
conf.level = 0.95,
...
)
variables_to_contrast(
model,
interactions = TRUE,
cross = FALSE,
var_categorical = "reference",
var_continuous = 1,
by_categorical = unique,
by_continuous = stats::fivenum
)Arguments
- x
(a model object, e.g.
glm)
A model to be tidied.- variables_list
(
listorstring)
A list whose elements will be sequentially passed tovariablesinmarginaleffects::avg_comparisons()(see details below); alternatively, it could also be the string"auto"(default),"cross"or"no_interaction"- conf.int
(
logical)
Whether or not to include a confidence interval in the tidied output.- conf.level
(
numeric)
The confidence level to use for the confidence interval (between0ans1).- ...
Additional parameters passed to
marginaleffects::avg_comparisons().- model
(a model object, e.g.
glm)
A model.- interactions
(
logical)
Should combinations of variables corresponding to interactions be returned?- cross
(
logical)
IfinteractionisTRUE, should "cross-contrasts" be computed? (ifFALSE, only the last term of an interaction is passed tovariableand the other terms are passed toby)- var_categorical
(
predictor values)
Defaultvariablevalue for categorical variables.- var_continuous
(
predictor values)
Defaultvariablevalue for continuous variables.- by_categorical
(
predictor values)
Defaultbyvalue for categorical variables.- by_continuous
(
predictor values)
Defaultbyvalue for continuous variables.
Details
Marginal contrasts are obtained by calling, for each variable or combination
of variables, marginaleffects::avg_comparisons().
tidy_marginal_contrasts() will compute marginal contrasts for each
variable or combination of variables, before stacking the results in a unique
tibble. This is why tidy_marginal_contrasts() has a variables_list
argument consisting of a list of specifications that will be passed
sequentially to the variables and the by argument of
marginaleffects::avg_comparisons().
Considering a single categorical variable named cat, tidy_marginal_contrasts()
will call avg_comparisons(model, variables = list(cat = "reference"))
to obtain average marginal contrasts for this variable.
Considering a single continuous variable named cont, tidy_marginalcontrasts()
will call avg_comparisons(model, variables = list(cont = 1))
to obtain average marginal contrasts for an increase of one unit.
For a combination of variables, there are several possibilities. You could
compute "cross-contrasts" by providing simultaneously several variables
to variables and specifying cross = TRUE to
marginaleffects::avg_comparisons(). Alternatively, you could compute the
contrasts of a first variable specified to variables for the
different values of a second variable specified to by.
The helper function variables_to_contrast() could be used to automatically
generate a suitable list to be used with variables_list. Each combination
of variables should be a list with two named elements: "variables" a list
of named elements passed to variables and "by" a list of named elements
used for creating a relevant datagrid and whose names are passed to by.
variables_list's default value, "auto", calls
variables_to_contrast(interactions = TRUE, cross = FALSE) while
"no_interaction" is a shortcut for
variables_to_contrast(interactions = FALSE). "cross" calls
variables_to_contrast(interactions = TRUE, cross = TRUE)
You can also provide custom specifications (see examples).
By default, average marginal contrasts are computed: contrasts are computed
using a counterfactual grid for each value of the variable of interest,
before averaging the results. Marginal contrasts at the mean could be
obtained by indicating newdata = "mean". Other assumptions are possible,
see the help file of marginaleffects::avg_comparisons().
For more information, see vignette("marginal_tidiers", "broom.helpers").
See also
marginaleffects::avg_comparisons(), tidy_avg_comparisons()
Other marginal_tieders:
tidy_all_effects(),
tidy_avg_comparisons(),
tidy_avg_slopes(),
tidy_ggpredict(),
tidy_marginal_predictions(),
tidy_margins()
Examples
# \donttest{
# Average Marginal Contrasts
df <- Titanic |>
dplyr::as_tibble() |>
tidyr::uncount(n) |>
dplyr::mutate(Survived = factor(Survived, c("No", "Yes")))
mod <- glm(
Survived ~ Class + Age + Sex,
data = df, family = binomial
)
tidy_marginal_contrasts(mod)
#> variable term estimate std.error statistic p.value
#> 1 Class 2nd - 1st -0.1951651 0.03654057 -5.341051 9.240906e-08
#> 2 Class 3rd - 1st -0.3066202 0.03069777 -9.988355 1.714025e-23
#> 3 Class Crew - 1st -0.1676466 0.03230024 -5.190258 2.100029e-07
#> 4 Age Child - Adult 0.1972150 0.04867887 4.051347 5.092362e-05
#> 5 Sex Male - Female -0.5073533 0.02452155 -20.690097 4.253562e-95
#> s.value conf.low conf.high predicted_lo predicted_hi predicted
#> 1 23.36739 -0.2667833 -0.1235469 NA NA NA
#> 2 75.62696 -0.3667867 -0.2464537 NA NA NA
#> 3 22.18309 -0.2309539 -0.1043393 NA NA NA
#> 4 14.26131 0.1018062 0.2926238 0.1039594 0.2511586 0.2511586
#> 5 313.49450 -0.5554146 -0.4592919 0.7904463 0.2511586 0.2511586
tidy_plus_plus(mod, tidy_fun = tidy_marginal_contrasts)
#> # A tibble: 5 × 23
#> term variable var_label var_class var_type var_nlevels contrasts
#> <chr> <chr> <chr> <chr> <chr> <int> <chr>
#> 1 2nd - 1st Class Class character categorical 4 contr.trea…
#> 2 3rd - 1st Class Class character categorical 4 contr.trea…
#> 3 Crew - 1st Class Class character categorical 4 contr.trea…
#> 4 Child - Adult Age Age character dichotomous 2 contr.trea…
#> 5 Male - Female Sex Sex character dichotomous 2 contr.trea…
#> # ℹ 16 more variables: contrasts_type <chr>, reference_row <lgl>, label <chr>,
#> # n_obs <dbl>, n_event <dbl>, estimate <dbl>, std.error <dbl>,
#> # statistic <dbl>, p.value <dbl>, s.value <dbl>, conf.low <dbl>,
#> # conf.high <dbl>, predicted_lo <dbl>, predicted_hi <dbl>, predicted <dbl>,
#> # label_attr <chr>
mod2 <- lm(Petal.Length ~ poly(Petal.Width, 2) + Species, data = iris)
tidy_marginal_contrasts(mod2)
#> variable term estimate std.error statistic p.value
#> 1 Petal.Width +1 0.9961779 0.1523954 6.536799 6.284962e-11
#> 2 Species versicolor - setosa 1.3170279 0.3145654 4.186817 2.828938e-05
#> 3 Species virginica - setosa 1.9123142 0.3733402 5.122176 3.020300e-07
#> s.value conf.low conf.high predicted_lo predicted_hi predicted
#> 1 33.88931 0.6974884 1.294867 1.390159 2.802537 1.390159
#> 2 15.10938 0.7004910 1.933565 NA NA NA
#> 3 21.65880 1.1805808 2.644048 NA NA NA
tidy_marginal_contrasts(
mod2,
variables_list = variables_to_predict(
mod2,
continuous = 3,
categorical = "pairwise"
)
)
#> variable term estimate std.error statistic p.value
#> 1 Petal.Width +3 1.7390991 1.0011046 1.737180 8.235537e-02
#> 2 Species versicolor - setosa 1.3170279 0.3145654 4.186817 2.828938e-05
#> 3 Species virginica - setosa 1.9123142 0.3733402 5.122176 3.020300e-07
#> 4 Species virginica - versicolor 0.5952863 0.1305792 4.558814 5.144328e-06
#> s.value conf.low conf.high predicted_lo predicted_hi predicted
#> 1 3.601993 -0.2230299 3.7012282 1.390159 4.377859 1.390159
#> 2 15.109380 0.7004910 1.9335648 NA NA NA
#> 3 21.658805 1.1805808 2.6440476 NA NA NA
#> 4 17.568586 0.3393558 0.8512168 NA NA NA
# Model with interactions
mod3 <- glm(
Survived ~ Sex * Age + Class,
data = df, family = binomial
)
tidy_marginal_contrasts(mod3)
#> variable term estimate std.error statistic p.value
#> 1 Class 2nd - 1st -0.19208546 0.03613415 -5.315898 1.061324e-07
#> 2 Class 3rd - 1st -0.30271279 0.03049293 -9.927311 3.166809e-23
#> 3 Class Crew - 1st -0.15351506 0.03217010 -4.771980 1.824240e-06
#> 4 Sex:Age Female * Child - Adult -0.02017115 0.06255589 -0.322450 7.471118e-01
#> 5 Sex:Age Male * Child - Adult 0.36915555 0.06129914 6.022198 1.720646e-09
#> s.value conf.low conf.high
#> 1 23.1676312 -0.2629071 -0.12126383
#> 2 74.7413161 -0.3624778 -0.24294775
#> 3 19.0642732 -0.2165673 -0.09046282
#> 4 0.4206039 -0.1427784 0.10243615
#> 5 29.1144025 0.2490114 0.48929966
tidy_marginal_contrasts(mod3, "no_interaction")
#> variable term estimate std.error statistic p.value
#> 1 Sex Male - Female -0.5216026 0.02386423 -21.857087 6.655897e-106
#> 2 Age Child - Adult 0.2891015 0.05091793 5.677793 1.364433e-08
#> 3 Class 2nd - 1st -0.1920855 0.03613415 -5.315898 1.061324e-07
#> 4 Class 3rd - 1st -0.3027128 0.03049293 -9.927311 3.166809e-23
#> 5 Class Crew - 1st -0.1535151 0.03217010 -4.771980 1.824240e-06
#> s.value conf.low conf.high predicted_lo predicted_hi predicted
#> 1 349.38974 -0.5683756 -0.47482955 0.56491193 0.3884185 0.3884185
#> 2 26.12712 0.1893042 0.38889880 0.09568894 0.3884185 0.3884185
#> 3 23.16763 -0.2629071 -0.12126383 NA NA NA
#> 4 74.74132 -0.3624778 -0.24294775 NA NA NA
#> 5 19.06427 -0.2165673 -0.09046282 NA NA NA
tidy_marginal_contrasts(mod3, "cross")
#> variable term estimate std.error statistic
#> 1 Class 2nd - 1st -0.1920855 0.03613415 -5.315898
#> 2 Class 3rd - 1st -0.3027128 0.03049293 -9.927311
#> 3 Class Crew - 1st -0.1535151 0.03217010 -4.771980
#> 4 Sex:Age Child - Adult * Male - Female -0.1695523 0.06280633 -2.699605
#> p.value s.value conf.low conf.high predicted_lo predicted_hi
#> 1 1.061324e-07 23.167631 -0.2629071 -0.12126383 NA NA
#> 2 3.166809e-23 74.741316 -0.3624778 -0.24294775 NA NA
#> 3 1.824240e-06 19.064273 -0.2165673 -0.09046282 NA NA
#> 4 6.942175e-03 7.170397 -0.2926505 -0.04645417 0.5917294 0.3884185
#> predicted
#> 1 NA
#> 2 NA
#> 3 NA
#> 4 0.3884185
tidy_marginal_contrasts(
mod3,
variables_list = list(
list(variables = list(Class = "pairwise"), by = list(Sex = unique)),
list(variables = list(Age = "all")),
list(variables = list(Class = "sequential", Sex = "reference"))
)
)
#> variable term estimate std.error statistic
#> 1 Sex:Class Female * 2nd - 1st -0.13997749 0.02991494 -4.679184
#> 2 Sex:Class Female * 3rd - 1st -0.30766670 0.02947081 -10.439711
#> 3 Sex:Class Female * 3rd - 2nd -0.16768921 0.03644975 -4.600559
#> 4 Sex:Class Female * Crew - 1st -0.10025576 0.02214882 -4.526461
#> 5 Sex:Class Female * Crew - 2nd 0.03972173 0.03075434 1.291581
#> 6 Sex:Class Female * Crew - 3rd 0.20741095 0.02966665 6.991383
#> 7 Sex:Class Male * 2nd - 1st -0.20622418 0.03884619 -5.308736
#> 8 Sex:Class Male * 3rd - 1st -0.30253894 0.03297582 -9.174568
#> 9 Sex:Class Male * 3rd - 2nd -0.09631477 0.02470330 -3.898862
#> 10 Sex:Class Male * Crew - 1st -0.16777963 0.03553556 -4.721458
#> 11 Sex:Class Male * Crew - 2nd 0.03844455 0.02847019 1.350344
#> 12 Sex:Class Male * Crew - 3rd 0.13475931 0.01866581 7.219578
#> 13 Age Adult - Child -0.28910149 0.05091793 -5.677793
#> 14 Age Child - Adult 0.28910149 0.05091793 5.677793
#> 15 Class:Sex 2nd - 1st * Male - Female -0.69156750 0.03221261 -21.468846
#> 16 Class:Sex 3rd - 2nd * Male - Female -0.64790478 0.03583072 -18.082385
#> 17 Class:Sex Crew - 3rd * Male - Female -0.34545626 0.03495969 -9.881560
#> p.value s.value conf.low conf.high
#> 1 2.880191e-06 18.405404 -0.19860969 -0.08134529
#> 2 1.633079e-25 82.340608 -0.36542842 -0.24990498
#> 3 4.213585e-06 17.856520 -0.23912940 -0.09624902
#> 4 5.997974e-06 17.347093 -0.14366664 -0.05684487
#> 5 1.965021e-01 2.347383 -0.02055566 0.09999913
#> 6 2.721895e-12 38.418525 0.14926537 0.26555652
#> 7 1.103879e-07 23.110914 -0.28236131 -0.13008704
#> 8 4.533874e-20 64.257818 -0.36717037 -0.23790752
#> 9 9.664587e-05 13.336932 -0.14473235 -0.04789718
#> 10 2.341602e-06 18.704073 -0.23742804 -0.09813122
#> 11 1.769057e-01 2.498948 -0.01735600 0.09424509
#> 12 5.214898e-13 40.802426 0.09817499 0.17134364
#> 13 1.364433e-08 26.127123 -0.38889880 -0.18930418
#> 14 1.364433e-08 26.127123 0.18930418 0.38889880
#> 15 3.044879e-102 337.230281 -0.75470306 -0.62843195
#> 16 4.386928e-73 240.367540 -0.71813169 -0.57767787
#> 17 5.004764e-23 74.081044 -0.41397599 -0.27693653
mod4 <- lm(Sepal.Length ~ Petal.Length * Petal.Width + Species, data = iris)
tidy_marginal_contrasts(mod4)
#> variable term estimate std.error
#> 1 Species versicolor - setosa -1.462933106 0.3719381
#> 2 Species virginica - setosa -1.984239711 0.4231549
#> 3 Petal.Length:Petal.Width 1 * +1 -0.125145768 0.3141612
#> 4 Petal.Length:Petal.Width 1.6 * +1 -0.105804839 0.2767317
#> 5 Petal.Length:Petal.Width 4.35 * +1 -0.017158912 0.1587605
#> 6 Petal.Length:Petal.Width 5.1 * +1 0.007017249 0.1594941
#> 7 Petal.Length:Petal.Width 6.9 * +1 0.065040037 0.2256260
#> statistic p.value s.value conf.low conf.high
#> 1 -3.93326994 8.379804e-05 13.54272399 -2.1919185 -0.7339477
#> 2 -4.68915671 2.743332e-06 18.47563910 -2.8136081 -1.1548713
#> 3 -0.39834888 6.903730e-01 0.53455198 -0.7408904 0.4905989
#> 4 -0.38233719 7.022113e-01 0.51002294 -0.6481891 0.4365794
#> 5 -0.10808052 9.139318e-01 0.12984155 -0.3283237 0.2940059
#> 6 0.04399693 9.649069e-01 0.05153841 -0.3055854 0.3196199
#> 7 0.28826483 7.731440e-01 0.37119090 -0.3771788 0.5072588
tidy_marginal_contrasts(
mod4,
variables_list = list(
list(
variables = list(Species = "sequential"),
by = list(Petal.Length = c(2, 5))
),
list(
variables = list(Petal.Length = 2),
by = list(Species = unique, Petal.Width = 2:4)
)
)
)
#> variable term estimate
#> 1 Petal.Length:Species 2 * versicolor - setosa -1.4629331
#> 2 Petal.Length:Species 2 * virginica - versicolor -0.5213066
#> 3 Petal.Length:Species 5 * versicolor - setosa -1.4629331
#> 4 Petal.Length:Species 5 * virginica - versicolor -0.5213066
#> 5 Species:Petal.Width:Petal.Length setosa * 2 * +2 1.8375120
#> 6 Species:Petal.Width:Petal.Length setosa * 3 * +2 1.9019817
#> 7 Species:Petal.Width:Petal.Length setosa * 4 * +2 1.9664515
#> 8 Species:Petal.Width:Petal.Length versicolor * 2 * +2 1.8375120
#> 9 Species:Petal.Width:Petal.Length versicolor * 3 * +2 1.9019817
#> 10 Species:Petal.Width:Petal.Length versicolor * 4 * +2 1.9664515
#> 11 Species:Petal.Width:Petal.Length virginica * 2 * +2 1.8375120
#> 12 Species:Petal.Width:Petal.Length virginica * 3 * +2 1.9019817
#> 13 Species:Petal.Width:Petal.Length virginica * 4 * +2 1.9664515
#> std.error statistic p.value s.value conf.low conf.high
#> 1 0.3719381 -3.933270 8.379805e-05 13.54272 -2.1919185 -0.7339477
#> 2 0.1263081 -4.127262 3.671076e-05 14.73344 -0.7688659 -0.2737473
#> 3 0.3719381 -3.933270 8.379793e-05 13.54273 -2.1919184 -0.7339478
#> 4 0.1263081 -4.127262 3.671076e-05 14.73344 -0.7688659 -0.2737473
#> 5 0.1601261 11.475403 1.753594e-30 98.84753 1.5236705 2.1513535
#> 6 0.2541642 7.483280 7.249039e-14 43.64920 1.4038291 2.4001344
#> 7 0.3833691 5.129395 2.906752e-07 21.71409 1.2150619 2.7178412
#> 8 0.1601261 11.475403 1.753595e-30 98.84753 1.5236705 2.1513535
#> 9 0.2541642 7.483280 7.249036e-14 43.64920 1.4038291 2.4001344
#> 10 0.3833691 5.129394 2.906758e-07 21.71409 1.2150618 2.7178412
#> 11 0.1601261 11.475403 1.753595e-30 98.84753 1.5236705 2.1513535
#> 12 0.2541642 7.483280 7.249033e-14 43.64920 1.4038291 2.4001344
#> 13 0.3833691 5.129394 2.906758e-07 21.71409 1.2150618 2.7178412
# Marginal Contrasts at the Mean
tidy_marginal_contrasts(mod, newdata = "mean")
#> variable term estimate std.error statistic p.value
#> 1 Class 2nd - 1st -0.2083189 0.03909115 -5.329054 9.872545e-08
#> 2 Class 3rd - 1st -0.3030788 0.03320759 -9.126792 7.055934e-20
#> 3 Class Crew - 1st -0.1815385 0.03558936 -5.100919 3.380076e-07
#> 4 Age Child - Adult 0.2315175 0.06082479 3.806301 1.410609e-04
#> 5 Sex Male - Female -0.5405541 0.02514911 -21.493967 1.772950e-102
#> s.value conf.low conf.high rowid predicted_lo predicted_hi predicted
#> 1 23.27200 -0.2849361 -0.1317016 NA NA NA NA
#> 2 63.61972 -0.3681645 -0.2379931 NA NA NA NA
#> 3 21.49644 -0.2512924 -0.1117846 NA NA NA NA
#> 4 12.79139 0.1123030 0.3507319 1 0.2254997 0.4570172 0.2254997
#> 5 338.01051 -0.5898454 -0.4912627 1 0.7660538 0.2254997 0.2254997
tidy_marginal_contrasts(mod3, newdata = "mean")
#> variable term estimate std.error statistic
#> 1 Class 2nd - 1st -0.20575641 0.03889922 -5.2894733
#> 2 Class 3rd - 1st -0.29710778 0.03330055 -8.9220082
#> 3 Class Crew - 1st -0.16816852 0.03583195 -4.6932563
#> 4 Sex:Age Female * Child - Adult -0.01826595 0.05710961 -0.3198401
#> 5 Sex:Age Male * Child - Adult 0.41025175 0.06569429 6.2448616
#> p.value s.value conf.low conf.high
#> 1 1.226691e-07 22.9587252 -0.2819975 -0.12951533
#> 2 4.579082e-19 60.9215753 -0.3623757 -0.23183990
#> 3 2.688903e-06 18.5045510 -0.2383978 -0.09793919
#> 4 7.490895e-01 0.4167899 -0.1301987 0.09366684
#> 5 4.241755e-10 31.1346196 0.2814933 0.53901020
# }