Artificial data for margins, copied from Stata

Source: R/margex.R

The dataset is identical to the one provided by Stata and available from webuse::webuse("margex") with categorical variables explicitly encoded as factors.

margex

Format

A data frame with 3000 observations on the following 11 variables.

y

A numeric vector

outcome

A binary numeric vector with values (0,1)

sex

A factor with two levels

group

A factor with three levels

age

A numeric vector

distance

A numeric vector

ycn

A numeric vector

yc

A numeric vector

treatment

A factor with two levels

agegroup

A factor with three levels

arm

A factor with three levels

Source

http://www.stata-press.com/data/r14/margex.dta

See also

prediction

Examples


# Examples from Stata's help files
# Also available from: webuse::webuse("margex")
data("margex")

# A simple case after regress
# . regress y i.sex i.group
# . margins sex
m1 <- lm(y ~ factor(sex) + factor(group), data = margex)
prediction(m1, at = list(sex = c("male", "female")))
#> Data frame with 6000 predictions from
#>  lm(formula = y ~ factor(sex) + factor(group), data = margex)
#> with average predictions:
#>     sex     x
#>    male 60.56
#>  female 78.88

# A simple case after logistic
# . logistic outcome i.sex i.group
# . margins sex
m2 <- glm(outcome ~ sex + group, binomial(), data = margex)
prediction(m2, at = list(sex = c("male", "female")))
#> Data frame with 6000 predictions from
#>  glm(formula = outcome ~ sex + group, family = binomial(), data = margex)
#> with average predictions:
#>     sex      x
#>    male 0.1287
#>  female 0.1905

# Average response versus response at average
# . margins sex
prediction(m2, at = list(sex = c("male", "female")))
#> Data frame with 6000 predictions from
#>  glm(formula = outcome ~ sex + group, family = binomial(), data = margex)
#> with average predictions:
#>     sex      x
#>    male 0.1287
#>  female 0.1905
# . margins sex, atmeans
## TODO

# Multiple margins from one margins command
# . margins sex group
prediction(m2, at = list(sex = c("male", "female")))
#> Data frame with 6000 predictions from
#>  glm(formula = outcome ~ sex + group, family = binomial(), data = margex)
#> with average predictions:
#>     sex      x
#>    male 0.1287
#>  female 0.1905
prediction(m2, at = list(group = c("1", "2", "3")))
#> Data frame with 9000 predictions from
#>  glm(formula = outcome ~ sex + group, family = binomial(), data = margex)
#> with average predictions:
#>  group       x
#>      1 0.28262
#>      2 0.10748
#>      3 0.02911

# Margins with interaction terms
# . logistic outcome i.sex i.group sex#group
# . margins sex group
m3 <- glm(outcome ~ sex * group, binomial(), data = margex)
prediction(m3, at = list(sex = c("male", "female")))
#> Data frame with 6000 predictions from
#>  glm(formula = outcome ~ sex * group, family = binomial(), data = margex)
#> with average predictions:
#>     sex      x
#>    male 0.1562
#>  female 0.1984
prediction(m3, at = list(group = c("1", "2", "3")))
#> Data frame with 9000 predictions from
#>  glm(formula = outcome ~ sex * group, family = binomial(), data = margex)
#> with average predictions:
#>  group      x
#>      1 0.3211
#>      2 0.1152
#>      3 0.0265

# Margins with continuous variables
# . logistic outcome i.sex i.group sex#group age
# . margins sex group
m4 <- glm(outcome ~ sex * group + age, binomial(), data = margex)
prediction(m4, at = list(sex = c("male", "female")))
#> Data frame with 6000 predictions from
#>  glm(formula = outcome ~ sex * group + age, family = binomial(), 
#>     data = margex)
#> with average predictions:
#>     sex      x
#>    male 0.1601
#>  female 0.1967
prediction(m4, at = list(group = c("1", "2", "3")))
#> Data frame with 9000 predictions from
#>  glm(formula = outcome ~ sex * group + age, family = binomial(), 
#>     data = margex)
#> with average predictions:
#>  group       x
#>      1 0.22513
#>      2 0.15060
#>      3 0.07362

# Margins of continuous variables
# . margins, at(age=40)
prediction(m4, at = list(age = 40))
#> Data frame with 3000 predictions from
#>  glm(formula = outcome ~ sex * group + age, family = binomial(), 
#>     data = margex)
#> with average prediction:
#>  age      x
#>   40 0.1134
# . margins, at(age=(30 35 40 45 50))
prediction(m4, at = list(age = c(30, 35, 40, 45, 50)))
#> Data frame with 15000 predictions from
#>  glm(formula = outcome ~ sex * group + age, family = binomial(), 
#>     data = margex)
#> with average predictions:
#>  age      x
#>   30 0.0500
#>   35 0.0760
#>   40 0.1134
#>   45 0.1647
#>   50 0.2315

# Margins of interactions
# . margins sex#group
prediction(m4, at = list(sex = c("male", "female"), group = c("1", "2", "3")))
#> Data frame with 18000 predictions from
#>  glm(formula = outcome ~ sex * group + age, family = binomial(), 
#>     data = margex)
#> with average predictions:
#>     sex group       x
#>    male     1 0.23796
#>  female     1 0.21586
#>    male     2 0.06583
#>  female     2 0.20544
#>    male     3 0.05380
#>  female     3 0.08545