Longitudinal data organization
Source:vignettes/longitudinal-data-organization.Rmd
longitudinal-data-organization.RmdLongitudinal data formats
Longitudinal data can be organized into two distinct formats:
- A person-level, wide, or multivariate, format where each person has only one row of data and multiple columns containing data from each measurement occasion.
glimpse(deviant_tolerance_pl)
#> Rows: 16
#> Columns: 8
#> $ id <fct> 9, 45, 268, 314, 442, 514, 569, 624, 723, 918, 949, 978, …
#> $ tolerance_11 <dbl> 2.23, 1.12, 1.45, 1.22, 1.45, 1.34, 1.79, 1.12, 1.22, 1.0…
#> $ tolerance_12 <dbl> 1.79, 1.45, 1.34, 1.22, 1.99, 1.67, 1.90, 1.12, 1.34, 1.0…
#> $ tolerance_13 <dbl> 1.90, 1.45, 1.99, 1.55, 1.45, 2.23, 1.90, 1.22, 1.12, 1.2…
#> $ tolerance_14 <dbl> 2.12, 1.45, 1.79, 1.12, 1.67, 2.12, 1.99, 1.12, 1.00, 1.9…
#> $ tolerance_15 <dbl> 2.66, 1.99, 1.34, 1.12, 1.90, 2.44, 1.99, 1.22, 1.12, 1.2…
#> $ male <dbl> 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0
#> $ exposure <dbl> 1.54, 1.16, 0.90, 0.81, 1.13, 0.90, 1.99, 0.98, 0.81, 1.2…- A person-period, long, or univariate, format where each person has one row of data for each measurement occasion.
glimpse(deviant_tolerance_pp)
#> Rows: 80
#> Columns: 5
#> $ id <fct> 9, 9, 9, 9, 9, 45, 45, 45, 45, 45, 268, 268, 268, 268, 268, …
#> $ age <dbl> 11, 12, 13, 14, 15, 11, 12, 13, 14, 15, 11, 12, 13, 14, 15, …
#> $ tolerance <dbl> 2.23, 1.79, 1.90, 2.12, 2.66, 1.12, 1.45, 1.45, 1.45, 1.99, …
#> $ male <dbl> 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, …
#> $ exposure <dbl> 1.54, 1.54, 1.54, 1.54, 1.54, 1.16, 1.16, 1.16, 1.16, 1.16, …Most R functions expect data to be in the person-period format for visualization and analysis, but it’s easy to convert a longitudinal data set from one format to the other.
Converting between formats
To convert a person-level data set to person-period format use
dplyr::pivot_longer():
pivot_longer(
deviant_tolerance_pl,
cols = starts_with("tolerance_"),
names_to = "age",
names_pattern = "([[:digit:]]+)",
names_transform = as.integer,
values_to = "tolerance"
)
#> # A tibble: 80 × 5
#> id male exposure age tolerance
#> <fct> <dbl> <dbl> <int> <dbl>
#> 1 9 0 1.54 11 2.23
#> 2 9 0 1.54 12 1.79
#> 3 9 0 1.54 13 1.9
#> 4 9 0 1.54 14 2.12
#> 5 9 0 1.54 15 2.66
#> 6 45 1 1.16 11 1.12
#> 7 45 1 1.16 12 1.45
#> 8 45 1 1.16 13 1.45
#> 9 45 1 1.16 14 1.45
#> 10 45 1 1.16 15 1.99
#> # ℹ 70 more rowsTo convert a person-period data set to person-level format use
dplyr::pivot_wider():
pivot_wider(
deviant_tolerance_pp,
names_from = age,
names_prefix = "tolerance_",
values_from = tolerance
)
#> # A tibble: 16 × 8
#> id male exposure tolerance_11 tolerance_12 tolerance_13 tolerance_14
#> <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 9 0 1.54 2.23 1.79 1.9 2.12
#> 2 45 1 1.16 1.12 1.45 1.45 1.45
#> 3 268 1 0.9 1.45 1.34 1.99 1.79
#> 4 314 0 0.81 1.22 1.22 1.55 1.12
#> 5 442 0 1.13 1.45 1.99 1.45 1.67
#> 6 514 1 0.9 1.34 1.67 2.23 2.12
#> 7 569 0 1.99 1.79 1.9 1.9 1.99
#> 8 624 1 0.98 1.12 1.12 1.22 1.12
#> 9 723 0 0.81 1.22 1.34 1.12 1
#> 10 918 0 1.21 1 1 1.22 1.99
#> 11 949 1 0.93 1.99 1.55 1.12 1.45
#> 12 978 1 1.59 1.22 1.34 2.12 3.46
#> 13 1105 1 1.38 1.34 1.9 1.99 1.9
#> 14 1542 0 1.44 1.22 1.22 1.99 1.79
#> 15 1552 0 1.04 1 1.12 2.23 1.55
#> 16 1653 0 1.25 1.11 1.11 1.34 1.55
#> # ℹ 1 more variable: tolerance_15 <dbl>Adding discrete time indicators to person-period data
To add discrete time indicators to a person-period data set first
create a temporary copy of the time variable and a column of ones, then
use dplyr::pivot_wider():
deviant_tolerance_pp |>
mutate(
temp_age = age,
temp_dummy = 1
) |>
pivot_wider(
names_from = temp_age,
names_prefix = "age_",
values_from = temp_dummy,
values_fill = 0
)
#> # A tibble: 80 × 10
#> id age tolerance male exposure age_11 age_12 age_13 age_14 age_15
#> <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 9 11 2.23 0 1.54 1 0 0 0 0
#> 2 9 12 1.79 0 1.54 0 1 0 0 0
#> 3 9 13 1.9 0 1.54 0 0 1 0 0
#> 4 9 14 2.12 0 1.54 0 0 0 1 0
#> 5 9 15 2.66 0 1.54 0 0 0 0 1
#> 6 45 11 1.12 1 1.16 1 0 0 0 0
#> 7 45 12 1.45 1 1.16 0 1 0 0 0
#> 8 45 13 1.45 1 1.16 0 0 1 0 0
#> 9 45 14 1.45 1 1.16 0 0 0 1 0
#> 10 45 15 1.99 1 1.16 0 0 0 0 1
#> # ℹ 70 more rowsAdding contiguous periods to person-level survival data
To add contiguous periods to person-level data use
dplyr::reframe():
first_sex |>
# In order to add the event indicator, the time variable needs a different
# name in the person-level data from the name we want to use in `reframe()`.
# This is a temporary variable so it doesn't matter what the name is.
rename(grades = grade) |>
group_by(id) |>
reframe(
grade = 1:max(grades),
event = if_else(grade == grades & censor == 0, 1, 0),
# To keep predictors from the person-level data, simply list them. If there
# are many predictors it might be more convenient to use
# `dplyr::left_join()` after `reframe()`.
parental_transition,
parental_antisociality
)
#> # A tibble: 1,902 × 5
#> id grade event parental_transition parental_antisociality
#> <fct> <int> <dbl> <dbl> <dbl>
#> 1 1 1 0 0 1.98
#> 2 1 2 0 0 1.98
#> 3 1 3 0 0 1.98
#> 4 1 4 0 0 1.98
#> 5 1 5 0 0 1.98
#> 6 1 6 0 0 1.98
#> 7 1 7 0 0 1.98
#> 8 1 8 0 0 1.98
#> 9 1 9 1 0 1.98
#> 10 2 1 0 1 -0.545
#> # ℹ 1,892 more rows