cdid user guide

Guide for cdid 1.3.0.0 (R). Recorded examples retain their stated versions.

On this page

Work through a first chained difference-in-differences fit, then prepare your panel, choose comparisons and adjustments, and interpret the results. The package overview on GitHub has a runnable first example and the source installation requirements.

The guide describes cdid 1.3.0.0. Numerical output on the article pages comes from recorded runs, and each page states the package versions it used. Analytical population values are labeled separately. The release notes list what each version added.

Installation

Install the CRAN release with install.packages("cdid"). Use packageVersion("cdid") to check your version. For the examples that require 1.3.0.0, follow the GitHub installation instructions if your installed version is older.

Get started

Start with the worked fit on a balanced panel. To use your own data, supply a unit identifier, a finite numeric date, a numeric outcome with missing values kept as NA, and a treatment column. For staggered adoption, the treatment column holds each unit’s first treatment date, with 0 for units never treated. Prepare your panel and unit table gives the input checklist. Investigate missingness and data structure shows how to inspect who is observed at each date and which comparisons have support.

Each page of the reading path ends with links to the previous and next steps and a collapsible list of all steps.

Reading path

The 22 steps below are numbered as on the pages themselves. The stage headings group them by the decision they support. Read Stage 1 in order. Later stages can be read when the decision arises.

Stage 1. Make a first fit and understand what it estimates

Step Page What you learn
1 Get started Main functions on a balanced panel: estimation, summaries, aggregation, plots and export.
2 Moving from did What stays familiar, how incomplete panels are treated, and why balanced results can differ and which to use.
3 Incomplete panels Which changes an observation pattern supplies and how chaining combines them, with a rotating panel example.
4 Identify effects when the observed units change composition When changes within units identify effects that comparisons of observed levels do not.
5 How to compare precision Matched targets and assumptions, covariance, and what standard errors can show.

Stage 2. Prepare your data

Step Page What you learn
6 Prepare your panel and unit table Identifiers, missing outcomes kept as NA, the unit table, changing covariates or weights, and known reference covariate rows with missing_reference.

Stage 3. Adjust for treatment

Step Page What you learn
7 Treatment adjustment How IPW, OR and DR make a cohort comparable with its controls, and the separate or pooled outcome regression.
8 Choose IPW, OR or DR in an incomplete panel The three treatment adjustments compared on one incomplete panel.

Stage 4. Handle missing outcomes

Step Page What you learn
9 Handling missing outcomes The study population, the assumption about missing outcomes and the three ways to adjust for them.
10 Sampling information for the observation model Advanced: fitting the observation model on part of the unit table.
11 Choose an observation model Date or pair logits, and how many slopes they share.
12 Impute actual changes Observation weighting, change imputation and the augmented correction compared on one panel.
Step Page What you learn
13 Parallel trends PT-post, PT-all and the separate reporting choice base_period.
14 Add PT-all restrictions What true and false PT-all restrictions change on one incomplete panel.
15 GMM weighting and covariance regularization How comparisons are combined and what covariance regularization does.
16 Robustness Which models must be correct for each combination of treatment adjustment and missing outcome handling.

Stage 6. Inference, reference and troubleshooting

Step Page What you learn
17 Inference Intervals, bands, clustering, aggregation and pretrend tests.
18 Options Every argument with its default, accepted values and when it is active.
19 Results The contents of each result, the uncertainty rules, pretrend tests and export.
20 Diagnose errors and unavailable results Refused requests, failed fits, unavailable effects and surprising results.

Stage 7. Chain across cohorts with different observation windows

Step Page What you learn
21 Cross-pair chaining, control parallel trends and within-unit weighting cross_pair = TRUE when a cohort and its controls share no date pair, pt_assumption = "ctrl", gmm_weighting = "within", and chain_report() to see which comparisons identify an effect.
22 Short pre-treatment windows Staggered cohorts observed only a few dates before adoption, and how chaining links their windows.

Choose the treatment design

The reading path uses the cohort model: gname holds each unit’s adoption date, treatment is absorbing, and effects may evolve with time since adoption. The switching model and the linear fixed-effects specification also support the treatment and missing-outcome adjustments of Stages 3 and 4.

Design Page What distinguishes it
Binary treatment that can stop and restart When treatment can stop and restart A static model without carryover: earlier treatment has no lingering effect beyond current status. Supply dname, the treatment status at every study date, including dates with missing outcomes. The result holds effects at entry and exit and a switcher-weighted average in fit$aggregate. aggte() does not apply.
Linear cohort-by-date fixed effects Cohort-by-date fixed effects A linear specification with unit fixed effects and a common untreated calendar path. Supply gname and control = list(preset = "cohort_time_fe"). The regression fit matches R etwfe and Stata jwdid on the same inputs.

Supplying both gname and dname is an error. Absorbing data supplied through dname still requests the static model and its targets. The input requirements of the switching model differ from the cohort route, and the options reference marks which arguments apply to each.

Compare with other software

Hands-on notebook

The practical notebook covers the main workflows with editable synthetic examples: a first balanced fit, missing-outcome corrections, treatment adjustment, parallel trends and cross-pair chains, inference, pretrend tests, robustness, dated covariates, GMM diagnostics, the switching model and the cohort-by-date preset. Use the editable Quarto source to change the configuration and rerun, or download the standalone HTML and open it in a browser to read the tables and plots without running R.

Which page answers my question?

Question Where to look
An effect is reported as unavailable An effect is unavailable and the available and support.reason columns in Tables from tidy and glance.
A covariate is missing at a cohort’s reference date Add a known reference covariate row. The default missing_reference = "exclude" warns and restricts the affected cohort, and "error" stops and lists the inputs.
Should I use chained or ordinary mode on a balanced panel Why balanced results can differ and which to use.
My treatment stops or restarts When treatment can stop and restart.
My cohort and its controls never share a date pair Cross-pair chaining, with chain_report() to inspect the comparisons that identify each effect.
Which models must be correct for my choice of adjustments Which models must be correct and Robustness.
I want to see the fitted equations and coefficients Set control = list(keep_nuisance = TRUE) and call nuisance_models(fit). Section 8 of the practical notebook and When treatment can stop and restart show the reports.
The pretrend test is unavailable Pretrend tests are unavailable.
I want the regression equivalent in R or Stata Match R etwfe and Stata jwdid and Reproduce R and Stata models.
A call stopped with a refusal Refusals lists the error classes, and Diagnose errors explains what to check.

Terms used in this guide

Term Meaning
Study population The population whose treatment effects the analysis describes.
Unit table One row per study unit, supplied through the frame argument, including units with one or no observed outcomes. Without frame, the units present in data form it.
Treatment cohort Units first treated in the same period. gname holds that period, and 0 marks units never treated.
Switchers and stayers On the dname route, units whose treatment status changes at a date and units with the same prior status that keep it. Stayers are the comparison for that transition.
Units observed at both dates The units whose outcomes are observed at both dates of a comparison. Only they contribute that change.
Treatment adjustment How each cohort is made comparable with its controls: IPW, OR or DR, chosen with est_method.
Handling missing outcomes The assumption made about missing outcomes and the way the package adjusts for them, chosen with the arguments that begin with selection and with imputation_pooling.
PT-post, PT-ctrl and PT-all Parallel trends from each cohort’s reference period onward, the same with common earlier trends among the clean control cohorts, and the same with common earlier trends for all cohorts from a chosen start date.

Questions and bug reports

Use Issues for questions about using the package or choosing a method and for reproducible bugs. Include the package version and a reproducible example.

See Contributing to suggest a change.

Support the project

If cdid is useful to your work, you can buy me a coffee.

View this page at its source version ยท Report a problem