Simulate Some Simple Example Data

This function is mostly used to obtain example data to illustrate the features of the SPMD package. It simulates a dataset with three main variables: X (a continuous time-fixed confounder), A (a binary time-dependent exposure variable) and Y (a recurrent event outcome). The data is created in the start-stop format, as required for the sym_pair_matching function. For more details on the data generation process, see details.

Usage

sim_example_data(n, rr=2.5, risk_period=40, max_t=1000)

Arguments

n: A single positive number, specifying the number of individuals that should be simulated.
rr: A single positive number, specifying the true effect of the exposure (A) on the outcome (Y) during the risk_period time units after exposure.
risk_period: A single positive integer, specifying the length of the risk period during which the outcome probability is changed after an exposure occurs.
max_t: A single positive number specifying the amount of points in time that should be generated.

Details

In this data, X is simulated first using draws from a standard normal distribution. The probability for exposure A turning TRUE is dependent on time and X. The probability for an event in Y is likewise dependent on time, X and whether the individual is currently in the risk period after having experienced an exposure A. Because X causes both A and Y, it is a confounder for the relationship between A and Y. Similarly, because the probability for both A and Y is time-dependent, time itself acts as a confounding variable. There are no time-dependent confounders and no interactions between X and A or time and A. For simplicity, an individual can only ever be exposed once.

The data is generated using a discrete-event simulation approach with a Gillespie type algorithm as implemented in the simDAG R package. If users want to generate similar data, we highly recommend checking out that package.

Author

Robin Denz

Value

Returns a data.table containing six columns: .id (the unique person identifier), start and stop (defining some durations of individual-specific time in which no variables changed), X (a continuous time-fixed confounder), A (a time-dependent logical exposure variable) and Y (a recurrent logical event).

References

Denz, Robin and Nina Timmesfeld (2026). "Simulating Complex Cross-Sectional and Longitudinal Data using the simDAG R Package". Journal of Statistical Software 116 (2), doi: 10.18637/jss.v116.i02.

Gillespie, Daniel T. 1976. “A General Method for Numerically Simulating the Stochastic Time Evolution of Coupled Chemical Reactions.” Journal of Computational Physics 22 (4): 403–34.

Denz, Robin, Filippo Saatkamp, Katharina Meiszl and Nina Timmesfeld (2026). "The Symmetric Pair Matching Design: A Self-Controlled Method with Automatic Adjustment for Time Effects". arXiv Preprint. doi: 10.48550/arXiv.2608.25979.

Examples

set.seed(1234)

# simulate data with no exposure effect
data <- sim_example_data(n=100, rr=1)
head(data)
#> Key: <.id, start>
#>      .id start  stop          X      A      Y
#>    <int> <num> <num>      <num> <lgcl> <lgcl>
#> 1:     1     0   386 -1.2070657  FALSE  FALSE
#> 2:     1   386   426 -1.2070657   TRUE  FALSE
#> 3:     1   426  1000 -1.2070657  FALSE  FALSE
#> 4:     2     0    86  0.2774292  FALSE   TRUE
#> 5:     2    86    87  0.2774292  FALSE  FALSE
#> 6:     2    87   188  0.2774292  FALSE  FALSE

# simulate data where the exposure increases the event probability
data <- sim_example_data(n=100, rr=2.5)
head(data)
#> Key: <.id, start>
#>      .id start  stop          X      A      Y
#>    <int> <num> <num>      <num> <lgcl> <lgcl>
#> 1:     1     0   984 -3.3960635  FALSE   TRUE
#> 2:     1   984   985 -3.3960635  FALSE  FALSE
#> 3:     1   985  1000 -3.3960635  FALSE  FALSE
#> 4:     2     0   302 -0.7813523  FALSE  FALSE
#> 5:     2   302   342 -0.7813523   TRUE  FALSE
#> 6:     2   342  1000 -0.7813523  FALSE  FALSE