exercise_data.RData

`exercise_data.RData`

The dataset for the nine exercise sections. Generated by exercises/simulate_exercise_data.R.

The exercise scenario studies how work-related psychological states of one partner crossover to influence the other partner’s work outcomes, and how dyad-level moderators shape those pathways.

Download exercise_data.RData{.btn .btn-primary download}

Object	Format	Dimensions	Description
`ddl2`	long	500 rows × 15 columns	One row per person
`ddw2`	wide	250 rows × 16 columns	One row per dyad

Variables

Person-level (in long format `ddl2`)

Variable	Type	Description
`dyad_id`	integer	Dyad identifier (1–250)
`person_id`	integer	Person within dyad (1 or 2)
`gender`	factor (`male`, `female`)	Distinguishing variable
`affect`	numeric	Positive work-related affect
`sdt`	numeric	Self-determination at work
`job_crafting`	numeric	Proactive job-crafting behaviour
`partner_affect`	numeric	Partner’s affect
`partner_sdt`	numeric	Partner’s self-determination
`partner_job_crafting`	numeric	Partner’s job-crafting
`engagement`	numeric	Work engagement
`performance`	numeric	In-role performance
`creativity`	numeric	Creative performance
`live_together`	numeric (0/1)	Dyad lives together
`years_together`	numeric	Years as a couple
`time_spent_this_morning_together`	numeric	Minutes together this morning

Wide-format equivalents (in `ddw2`)

Each person-level variable has an _a and _p version. Dyad-level moderators appear once.

Design features

Three outcomes, three crossover pathways

Each outcome has one focal partner predictor and one focal dyad-level moderator. The crossover effects are:

Outcome	Focal partner predictor	Focal moderator	DGP interaction
`engagement`	`partner_job_crafting`	`live_together`	partner_jc × live_together = 0.15
`performance`	`partner_sdt`	`years_together`	partner_sdt × years_together = 0.015
`creativity`	`partner_affect`	`time_spent_this_morning_together`	partner_affect × time = 0.003

These focal interactions are small but detectable at N = 250 couples (500 individuals). Your likelihood ratio tests should recover them.

Within-dyad coupling

Variable pair	Within-dyad correlation
`affect` (predictor)	0.30
`sdt` (predictor)	0.30
`job_crafting` (predictor)	0.30
`engagement` (outcome residual)	0.25
`performance` (outcome residual)	0.25
`creativity` (outcome residual)	0.25

One non-zero gender interaction in the DGP

Out of all possible gender × predictor interactions, only one is non-zero in the data-generating process: actor_sdt × gender_male on engagement. This is the only interaction the gender-moderator LRTs should recover. The exercise 4 section asks you to check this.

Why the dataset is different from `dyad_data.RData`

The exercise dataset has three outcomes (not one) and three moderators (not one). This forces you to repeat your analysis pipeline for each combination, which is what you would do in a real research project. The repetition is the point.

The script number convention changes: ddl2 and ddw2 instead of ddl and ddw. This avoids name collisions when both files are loaded in the same R session.

Reproducibility

set.seed(2026) is set at the top of the simulation script. Re-running exercises/simulate_exercise_data.R will always produce the same RData file.

Loading

load("data/exercise_data.RData")
ls()                 # "ddl2"  "ddw2"
head(ddl2, 3)
head(ddw2, 3)

--- title: "exercise_data.RData" --- # `exercise_data.RData` The dataset for the nine [exercise sections](../exercises/index.html). Generated by [`exercises/simulate_exercise_data.R`](https://github.com/michaelides/apim/blob/master/exercises/simulate_exercise_data.R). The exercise scenario studies how work-related psychological states of one partner **crossover** to influence the other partner's work outcomes, and how dyad-level moderators shape those pathways. [Download exercise_data.RData](https://github.com/michaelides/apim/raw/master/data/exercise_data.RData){.btn .btn-primary download} ## Contents | Object | Format | Dimensions | Description | |---|---|---|---| | `ddl2` | long | 500 rows × 15 columns | One row per person | | `ddw2` | wide | 250 rows × 16 columns | One row per dyad | ## Variables ### Person-level (in long format `ddl2`) | Variable | Type | Description | |---|---|---| | `dyad_id` | integer | Dyad identifier (1–250) | | `person_id` | integer | Person within dyad (1 or 2) | | `gender` | factor (`male`, `female`) | Distinguishing variable | | `affect` | numeric | Positive work-related affect | | `sdt` | numeric | Self-determination at work | | `job_crafting` | numeric | Proactive job-crafting behaviour | | `partner_affect` | numeric | Partner's affect | | `partner_sdt` | numeric | Partner's self-determination | | `partner_job_crafting` | numeric | Partner's job-crafting | | `engagement` | numeric | Work engagement | | `performance` | numeric | In-role performance | | `creativity` | numeric | Creative performance | | `live_together` | numeric (0/1) | Dyad lives together | | `years_together` | numeric | Years as a couple | | `time_spent_this_morning_together` | numeric | Minutes together this morning | ### Wide-format equivalents (in `ddw2`) Each person-level variable has an `_a` and `_p` version. Dyad-level moderators appear once. ## Design features ### Three outcomes, three crossover pathways Each outcome has one focal partner predictor and one focal dyad-level moderator. The crossover effects are: | Outcome | Focal partner predictor | Focal moderator | DGP interaction | |---|---|---|---| | `engagement` | `partner_job_crafting` | `live_together` | partner_jc × live_together = 0.15 | | `performance` | `partner_sdt` | `years_together` | partner_sdt × years_together = 0.015 | | `creativity` | `partner_affect` | `time_spent_this_morning_together` | partner_affect × time = 0.003 | These focal interactions are small but detectable at N = 250 couples (500 individuals). Your likelihood ratio tests should recover them. ### Within-dyad coupling | Variable pair | Within-dyad correlation | |---|---| | `affect` (predictor) | 0.30 | | `sdt` (predictor) | 0.30 | | `job_crafting` (predictor) | 0.30 | | `engagement` (outcome residual) | 0.25 | | `performance` (outcome residual) | 0.25 | | `creativity` (outcome residual) | 0.25 | ### One non-zero gender interaction in the DGP Out of all possible gender × predictor interactions, only one is non-zero in the data-generating process: `actor_sdt × gender_male` on engagement. This is the only interaction the gender-moderator LRTs should recover. The [exercise 4 section](../exercises/section-4.html) asks you to check this. ## Why the dataset is different from `dyad_data.RData` The exercise dataset has *three* outcomes (not one) and *three* moderators (not one). This forces you to repeat your analysis pipeline for each combination, which is what you would do in a real research project. The repetition is the point. The script number convention changes: `ddl2` and `ddw2` instead of `ddl` and `ddw`. This avoids name collisions when both files are loaded in the same R session. ## Reproducibility `set.seed(2026)` is set at the top of the simulation script. Re-running `exercises/simulate_exercise_data.R` will always produce the same `RData` file. ## Loading ```r load("data/exercise_data.RData") ls() # "ddl2" "ddw2" head(ddl2, 3) head(ddw2, 3) ```