Variable types: rowdep vs rowind • swereg

library(swereg)
#> swereg 26.4.2
#> https://papadopoulos-lab.github.io/swereg/
library(data.table)

Introduction

In longitudinal registry data analysis with swereg, variables can be classified into two fundamental types:

rowdep (row-dependent): Variables that can change over time for a person
rowind (row-independent): Variables that cannot change over time for a person

Understanding this distinction is crucial for effective analysis of Swedish registry data, particularly during the skeleton2_clean phase where many transformations convert rowdep variables into rowind variables.

The concept

Row-dependent (rowdep) variables

These variables can have different values across time periods for the same person:

Education level (rowdep_edu_cat): A person’s education can improve over time
Income (rowdep_income_inflation_adjusted): Income changes annually
Had diagnosis this week (diag_gd_icd10_F64_089): TRUE/FALSE depending on the specific week
Current age (age): Increases continuously over time

Row-independent (rowind) variables

These variables have the same value across all time periods for a person:

Age at first diagnosis (rowind_age_first_gd): Fixed once the diagnosis occurs
Year of first diagnosis (rowind_isoyear_first_gd): Fixed historical fact
Birth country (rowind_birthcountry): Never changes
Register classification (rowind_register_tag): Person’s role in the study (case, control, etc.)
Age at death (rowind_age_death): Fixed once death occurs

Why this matters

In longitudinal data, you often need to:

Identify when something first happened (e.g., first diagnosis, first prescription)
Capture characteristics at specific time points (e.g., education level when first diagnosed)
Create person-level summaries that don’t vary by time period

Converting rowdep → rowind allows you to create stable, person-level characteristics for analysis.

Common transformation patterns

Pattern 1: First occurrence transformations

The most common transformation finds the first time a condition is TRUE and extracts a value from that time point. This works for any condition - time-based, value-based, or complex combinations.

Time-based conditions

# Age at first GD diagnosis for AFAB individuals
make_rowind_first_occurrence(skeleton_gd,
                            condition = "diag_gd_icd10_F64_089 == TRUE & is_amab == FALSE",
                            value_var = "age", 
                            new_var = "rowind_age_first_gd_afab")

# Education level at time of first diagnosis
make_rowind_first_occurrence(skeleton_gd,
                            condition = "isoyearweek == rowind_isoyearweek_first_gd",
                            value_var = "rowdep_edu_cat", 
                            new_var = "rowind_edu_at_first_dx")

Manual approach (for understanding)

# The helper function does this automatically:
skeleton_gd[diag_gd_icd10_F64_089 == TRUE & is_amab == FALSE, temp := age]
skeleton_gd[, rowind_age_first_gd_afab := swereg::first_non_na(temp), by = .(id)]
skeleton_gd[, temp := NULL]  # Always clean up temp variable

Pattern 2: Simple renaming (already rowind)

When variables are already row-independent but need consistent naming:

# If date of birth is truly the same for all rows (as it should be)
setnames(skeleton_gd, "dob", "rowind_dob")

# If birth country was added from demographics (already person-level)
setnames(skeleton_gd, "birth_country", "rowind_birth_country")

Important: Only use this when the variable is genuinely the same across all rows for each person. If values differ, you have a data integration problem to fix first.

Practical example with fake data

Let’s demonstrate these concepts using the fake datasets included in swereg:

# Create a small skeleton for demonstration
ids <- swereg::fake_demographics$lopnr[1:5]
skeleton <- create_skeleton(ids, "2020-01-01", "2020-03-31")

# Add demographic data (creates rowind variables)
fake_demographics <- swereg::fake_demographics |>
  data.table::copy() |>
  swereg::make_lowercase_names(date_columns = "doddatum")
#> Found additional date columns not in date_columns: fodelseman. Consider adding them for automatic date parsing.
add_onetime(skeleton, fake_demographics, "lopnr")

# Add diagnosis data (creates rowdep variables)
fake_diagnoses <- swereg::fake_diagnoses |>
  data.table::copy() |>
  swereg::make_lowercase_names(date_columns = "indatum")
#> Found additional date columns not in date_columns: utdatum. Consider adding them for automatic date parsing.
add_diagnoses(skeleton, fake_diagnoses, "lopnr", 
             diags = list("f64_diag" = "^F64"))
#> Warning: 'diags' is deprecated, use 'codes' instead.

# Examine the structure
print("Skeleton structure:")
#> [1] "Skeleton structure:"
head(skeleton[id == ids[1]], 8)
#>       id isoyear isoyearweek is_isoyear isoyearweeksun personyears fodelseman
#>    <int>   <int>      <char>     <lgcl>         <Date>       <num>     <char>
#> 1:     1    1900     1900-**       TRUE     1900-07-01           1       1959
#> 2:     1    1901     1901-**       TRUE     1901-06-30           1       1959
#> 3:     1    1902     1902-**       TRUE     1902-06-29           1       1959
#> 4:     1    1903     1903-**       TRUE     1903-06-28           1       1959
#> 5:     1    1904     1904-**       TRUE     1904-07-03           1       1959
#> 6:     1    1905     1905-**       TRUE     1905-07-02           1       1959
#> 7:     1    1906     1906-**       TRUE     1906-07-01           1       1959
#> 8:     1    1907     1907-**       TRUE     1907-06-30           1       1959
#>    doddatum f64_diag
#>      <Date>   <lgcl>
#> 1:     <NA>    FALSE
#> 2:     <NA>    FALSE
#> 3:     <NA>    FALSE
#> 4:     <NA>    FALSE
#> 5:     <NA>    FALSE
#> 6:     <NA>    FALSE
#> 7:     <NA>    FALSE
#> 8:     <NA>    FALSE

Now let’s create rowind variables:

# Example 1: ISO year of first F64 diagnosis  
make_rowind_first_occurrence(skeleton,
                            condition = "f64_diag == TRUE",
                            value_var = "isoyear", 
                            new_var = "rowind_isoyear_first_f64")

# Example 2: ISO year-week of first F64 diagnosis
make_rowind_first_occurrence(skeleton,
                            condition = "f64_diag == TRUE",
                            value_var = "isoyearweek", 
                            new_var = "rowind_isoyearweek_first_f64")

# View results
print("Results with rowind variables:")
#> [1] "Results with rowind variables:"
head(skeleton[id == ids[1], .(id, isoyear, isoyearweek, f64_diag, 
                              rowind_isoyear_first_f64, rowind_isoyearweek_first_f64)], 8)
#>       id isoyear isoyearweek f64_diag rowind_isoyear_first_f64
#>    <int>   <int>      <char>   <lgcl>                    <int>
#> 1:     1    1900     1900-**    FALSE                       NA
#> 2:     1    1901     1901-**    FALSE                       NA
#> 3:     1    1902     1902-**    FALSE                       NA
#> 4:     1    1903     1903-**    FALSE                       NA
#> 5:     1    1904     1904-**    FALSE                       NA
#> 6:     1    1905     1905-**    FALSE                       NA
#> 7:     1    1906     1906-**    FALSE                       NA
#> 8:     1    1907     1907-**    FALSE                       NA
#>    rowind_isoyearweek_first_f64
#>                          <char>
#> 1:                         <NA>
#> 2:                         <NA>
#> 3:                         <NA>
#> 4:                         <NA>
#> 5:                         <NA>
#> 6:                         <NA>
#> 7:                         <NA>
#> 8:                         <NA>

Notice how: - f64_diag varies by week (rowdep) - rowind_isoyear_first_f64 and rowind_isoyearweek_first_f64 are constant for each person (rowind)

Best practices

1. Naming conventions

Use clear prefixes to distinguish variable types: - rowdep_* for time-varying variables - rowind_* for time-invariant variables

2. Always clean up temporary variables

When using manual transformations, always remove temp variables:

skeleton[condition == TRUE, temp := value]
skeleton[, new_rowind_var := first_non_na(temp), by = .(id)]
skeleton[, temp := NULL]  # Critical: clean up

3. Use helper functions when possible

The make_rowind_first_occurrence() function handles temp variable management automatically and reduces errors.

4. Validate your transformations

Always check that your rowind variables are actually row-independent:

# This should return 1 for all persons (meaning all rows have same value)
skeleton[, .(unique_values = uniqueN(rowind_isoyear_first_f64)), by = .(id)]

Integration with swereg workflow

The rowdep/rowind concept fits into the standard swereg workflow:

skeleton1_create: Focus on data integration, creates mostly rowdep variables
skeleton2_clean: Heavy focus on rowdep → rowind transformations
skeleton3_analyze: Work with clean rowind variables for analysis

The skeleton2_clean phase is where most rowdep → rowind transformations occur, as you prepare stable person-level characteristics for downstream analysis.

Conclusion

Understanding the distinction between rowdep and rowind variables is fundamental to effective longitudinal registry data analysis. The transformation patterns shown here, especially when combined with helper functions like make_rowind_first_occurrence(), make it easier to create robust, analysis-ready datasets from complex Swedish registry data.

For more advanced workflows, see the other swereg vignettes on skeleton creation, data integration, and memory-efficient processing.