2 Outcome Trajectories

This chapter shows how to plot raw outcome variables as time series, colored by treatment status, using type = "outcome". The syntax is the same as for type = "treat" with type = "outcome" added.

2.1 Basic outcome plot

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"),
          type = "outcome", main = "EDR Reform and Turnout",
          ylim = c(0, 100), xlab = "Year", ylab = "Turnout")

Turn off pre/post coloring with pre.post = FALSE.

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"), pre.post = FALSE,
          type = "outcome", main = "EDR Reform and Turnout",
          ylim = c(0, 100), xlab = "Year", ylab = "Turnout")

Customise the legend, or suppress it entirely with legendOff = TRUE.

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"),
          type = "outcome", main = "EDR Reform and Turnout",
          legend.labs = c("Control States",
                          "Treated States (before EDR)",
                          "Treated States (after EDR)"))

Customise line colors and legend labels simultaneously.

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"),
          type = "outcome", main = "EDR Reform and Turnout",
          color = c("lightblue", "blue", "grey75"),
          legend.labs = c("Control States",
                          "Treated States (before EDR)",
                          "Treated States (after EDR)"))
#> Specified colors in the order of "treated (pre)", "treated (post)", "control".

2.2 Themes and overlays

Activate the high-contrast red theme for a publication-paper look. Under theme = "red", controls fade to a light gray, treated trajectories pick up a brick-red accent, and the treatment-onset marker becomes a solid black dashed line.

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"), type = "outcome",
          main = "EDR Reform and Turnout", theme = "red")

Overlay heavy group-mean trajectories and a 10–90% quantile ribbon with group.mean.overlay = TRUE. Useful when there are many units per group and you want the audience to see both individual variation and the group average at a glance. (Currently implemented for the main DID continuous- outcome path; combine with theme = "red" for the cleanest read.)

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"), type = "outcome",
          main = "EDR Reform and Turnout",
          theme = "red", group.mean.overlay = TRUE)

2.3 Plotting a subset of units

Use id to plot only specific units.

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"),
          type = "outcome", main = "EDR Reform and Turnout (AL, AR, CT)",
          id = c("AL", "AR", "CT"))

2.4 Splitting by treatment-status group

by.group = TRUE separates units into groups based on whether their treatment status ever changed (e.g., always treated, always control, switchers).

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"),
          type = "outcome", main = "EDR Reform and Turnout",
          by.group = TRUE)

Use by.group.side = TRUE to arrange the subfigures in a row rather than a column.

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"),
          type = "outcome", main = "EDR Reform and Turnout",
          by.group.side = TRUE)

2.5 Cohort-averaged trajectories

by.cohort = TRUE collapses units with identical treatment histories into cohort-average trajectories. This helps diagnose treatment-effect heterogeneity that may bias TWFE estimates.

panelview(turnout ~ policy_edr + policy_mail_in + policy_motor,
          data = turnout, index = c("abb", "year"),
          type = "outcome", main = "EDR Reform and Turnout",
          by.cohort = TRUE)
#> Number of unique treatment histories: 5

A more detailed cohort plot with custom legend labels and y-axis limits.

panelview(turnout ~ policy_edr + policy_motor,
          data = turnout, index = c("abb", "year"), type = "outcome",
          main = "EDR Reform and Turnout", by.cohort = TRUE,
          ylim = c(40, 80),
          legend.labs = c("Control States",
                          "Treated States (before EDR)",
                          "Treated States (after EDR)"))
#> Number of unique treatment histories: 5

2.6 Ignoring treatment status

Since v1.0.3, omitting the treatment indicator (using Y ~ 1) lets panelView visualize any variable in a panel dataset without reference to treatment.

panelview(turnout ~ 1, data = turnout, index = c("abb", "year"),
          type = "outcome", main = "Turnout",
          ylim = c(0, 100), xlab = "Year", ylab = "Turnout")

Equivalently, supply the variable name via Y =.

panelview(Y = "turnout", data = turnout,
          index = c("abb", "year"), type = "outcome",
          main = "Turnout", ylim = c(0, 100),
          xlab = "Year", ylab = "Turnout")

2.7 Treatment reversal and missing data

How panelView handles two common discontinuities in the outcome trajectory:

Treatment reversal (unit goes on then off, or off then on). The treated overlay is drawn only at periods the unit is treated, so each treated spell is a separate line segment.
Missing outcome. Under the default leave.gap = FALSE, rows with missing Y are dropped and the bracketing observed points are joined by a dotted segment so the gap is visible. Under leave.gap = TRUE the panel is expanded and the line breaks cleanly at every NA cell.

Topology classification (reversal vs. staggered) follows the observed data: if a unit’s leading treated rows are missing and the surviving D pattern is monotonic, the unit is plotted in the staggered palette.

Examples below use fect::sim_base. Unit 102 is treated at $t=1\!-\!3$, control at $t=4\!-\!26$, then treated at $t=27\!-\!35$.

2.7.1 Reversal alone

if (requireNamespace("fect", quietly = TRUE)) {
  data("sim_base", package = "fect")
  panelview(Y ~ D, data = subset(sim_base, id == 102),
            index = c("id", "time"), type = "outcome",
            main = "Unit 102: treated 1-3, control 4-26, treated 27-35")
}

2.7.2 Interior gap: dotted bridge under `leave.gap = FALSE`

Set $Y$ to NA at $t=10\!-\!12$ and $t=30\!-\!31$. The control line shows a dotted bridge between $t=9$ and $t=13$; the treated line shows one between $t=29$ and $t=32$.

if (requireNamespace("fect", quietly = TRUE)) {
  d <- subset(sim_base, id == 102)
  d$Y[d$time %in% c(10, 11, 12)] <- NA
  d$Y[d$time %in% c(30, 31)]     <- NA
  panelview(Y ~ D, data = d, index = c("id", "time"),
            type = "outcome",
            main = "leave.gap = FALSE: dotted bridge across missing periods")
}

2.7.3 Interior gap: clean break under `leave.gap = TRUE`

Same data, but leave.gap = TRUE keeps the NA rows, so the line breaks at every gap with no bridge.

if (requireNamespace("fect", quietly = TRUE)) {
  d <- subset(sim_base, id == 102)
  d$Y[d$time %in% c(10, 11, 12)] <- NA
  d$Y[d$time %in% c(30, 31)]     <- NA
  panelview(Y ~ D, data = d, index = c("id", "time"),
            type = "outcome", leave.gap = TRUE,
            main = "leave.gap = TRUE: line breaks at gaps, no bridge")
}

2.7.4 Missing at the boundary

Drop unit 102’s $Y$ at $t=1\!-\!3$ (the start of its series) and pass leave.gap = TRUE so the row is retained, the x-axis stays at $t=1\!-\!35$, and the topology classifier still sees the early treated $D$ values. The line is empty at $t=1\!-\!3$; everything else is unchanged.

if (requireNamespace("fect", quietly = TRUE)) {
  d <- subset(sim_base, id == 102)
  d$Y[d$time %in% 1:3] <- NA
  panelview(Y ~ D, data = d, index = c("id", "time"),
            type = "outcome", leave.gap = TRUE,
            main = "Drop Y at t=1-3 (leave.gap = TRUE keeps x-axis and classification)")
}

Under the default leave.gap = FALSE, the same NA injection drops those rows entirely. The panel for this unit shrinks to $t=4\!-\!35$, its surviving $D$ pattern becomes monotonic, and the classifier re-types it as staggered. Use leave.gap = TRUE whenever you need the boundary intact.

2.7.5 `by.group = TRUE` with all three subpanels

by.group = TRUE splits units by their treatment history: always under control, always under treatment, and treatment status changed. To exercise all three subpanels we promote one sim_base unit to always-treated.

if (requireNamespace("fect", quietly = TRUE)) {
  d <- sim_base
  d$D[d$id == 101] <- 1    ## promote unit 101 to always-treated
  panelview(Y ~ D, data = d, index = c("id", "time"),
            type = "outcome", by.group = TRUE,
            main = "by.group = TRUE: always control / always treated / status changed")
}

2.8 Discrete outcomes

Set outcome.type = "discrete" when the outcome takes a small number of integer values.

panelview(Y ~ D, data = simdata, index = c("id", "time"),
          outcome.type = "discrete",
          type = "outcome", xlim = c(8, 15))

Split by treatment-status group.

panelview(Y ~ D, data = simdata, index = c("id", "time"),
          by.group = TRUE, outcome.type = "discrete",
          type = "outcome", xlim = c(8, 15))

2.9 Multi-level or continuous treatment

When the treatment variable has more than 5 unique values (e.g., polity2), treatment status is not shown on the outcome plot.

panelview(Capacity ~ polity2 + lngdp,
          data = capacity, index = c("ccode", "year"),
          main = "Measuring State Capacity",
          type = "outcome", legendOff = TRUE)
#> 21 treatment levels.

# Outcome Trajectories {#sec-outcome} ```{r setup, include=FALSE} knitr::opts_chunk$set(echo = TRUE, message = FALSE, warning = FALSE, fig.width = 10, fig.height = 6) library(panelView) data(panelView) ``` This chapter shows how to plot raw outcome variables as time series, colored by treatment status, using `type = "outcome"`. The syntax is the same as for `type = "treat"` with `type = "outcome"` added. ## Basic outcome plot ```{r outcome1-1} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout", ylim = c(0, 100), xlab = "Year", ylab = "Turnout") ``` Turn off pre/post coloring with `pre.post = FALSE`. ```{r outcome1-1a} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), pre.post = FALSE, type = "outcome", main = "EDR Reform and Turnout", ylim = c(0, 100), xlab = "Year", ylab = "Turnout") ``` Customise the legend, or suppress it entirely with `legendOff = TRUE`. ```{r outcome1-2} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout", legend.labs = c("Control States", "Treated States (before EDR)", "Treated States (after EDR)")) ``` Customise line colors and legend labels simultaneously. ```{r outcome1-4} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout", color = c("lightblue", "blue", "grey75"), legend.labs = c("Control States", "Treated States (before EDR)", "Treated States (after EDR)")) ``` ## Themes and overlays Activate the high-contrast `red` theme for a publication-paper look. Under `theme = "red"`, controls fade to a light gray, treated trajectories pick up a brick-red accent, and the treatment-onset marker becomes a solid black dashed line. ```{r outcome1-3} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout", theme = "red") ``` Overlay heavy group-mean trajectories and a 10--90% quantile ribbon with `group.mean.overlay = TRUE`. Useful when there are many units per group and you want the audience to see both individual variation and the group average at a glance. (Currently implemented for the main DID continuous- outcome path; combine with `theme = "red"` for the cleanest read.) ```{r outcome1-3b} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout", theme = "red", group.mean.overlay = TRUE) ``` ## Plotting a subset of units Use `id` to plot only specific units. ```{r outcome1-5} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout (AL, AR, CT)", id = c("AL", "AR", "CT")) ``` ## Splitting by treatment-status group `by.group = TRUE` separates units into groups based on whether their treatment status ever changed (e.g., always treated, always control, switchers). ```{r outcome1-6} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout", by.group = TRUE) ``` Use `by.group.side = TRUE` to arrange the subfigures in a row rather than a column. ```{r outcome1-7} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout", by.group.side = TRUE) ``` ## Cohort-averaged trajectories `by.cohort = TRUE` collapses units with identical treatment histories into cohort-average trajectories. This helps diagnose treatment-effect heterogeneity that may bias TWFE estimates. ```{r outcome1-4a} panelview(turnout ~ policy_edr + policy_mail_in + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout", by.cohort = TRUE) ``` A more detailed cohort plot with custom legend labels and y-axis limits. ```{r outcome4-1} panelview(turnout ~ policy_edr + policy_motor, data = turnout, index = c("abb", "year"), type = "outcome", main = "EDR Reform and Turnout", by.cohort = TRUE, ylim = c(40, 80), legend.labs = c("Control States", "Treated States (before EDR)", "Treated States (after EDR)")) ``` ## Ignoring treatment status Since v1.0.3, omitting the treatment indicator (using `Y ~ 1`) lets **panelView** visualize any variable in a panel dataset without reference to treatment. ```{r outcome2-1} panelview(turnout ~ 1, data = turnout, index = c("abb", "year"), type = "outcome", main = "Turnout", ylim = c(0, 100), xlab = "Year", ylab = "Turnout") ``` Equivalently, supply the variable name via `Y =`. ```{r outcome2-2, eval=FALSE} panelview(Y = "turnout", data = turnout, index = c("abb", "year"), type = "outcome", main = "Turnout", ylim = c(0, 100), xlab = "Year", ylab = "Turnout") ``` ## Treatment reversal and missing data {#sec-outcome-reversal} How **panelView** handles two common discontinuities in the outcome trajectory: - **Treatment reversal** (unit goes on then off, or off then on). The treated overlay is drawn only at periods the unit is treated, so each treated spell is a separate line segment. - **Missing outcome**. Under the default `leave.gap = FALSE`, rows with missing `Y` are dropped and the bracketing observed points are joined by a **dotted** segment so the gap is visible. Under `leave.gap = TRUE` the panel is expanded and the line breaks cleanly at every NA cell. Topology classification (reversal vs. staggered) follows the observed data: if a unit's leading treated rows are missing and the surviving `D` pattern is monotonic, the unit is plotted in the staggered palette. Examples below use `fect::sim_base`. Unit 102 is treated at $t=1\!-\!3$, control at $t=4\!-\!26$, then treated at $t=27\!-\!35$. ### Reversal alone ```{r outcome-reversal, fig.height=4.2} if (requireNamespace("fect", quietly = TRUE)) { data("sim_base", package = "fect") panelview(Y ~ D, data = subset(sim_base, id == 102), index = c("id", "time"), type = "outcome", main = "Unit 102: treated 1-3, control 4-26, treated 27-35") } ``` ### Interior gap: dotted bridge under `leave.gap = FALSE` Set $Y$ to `NA` at $t=10\!-\!12$ and $t=30\!-\!31$. The control line shows a dotted bridge between $t=9$ and $t=13$; the treated line shows one between $t=29$ and $t=32$. ```{r outcome-missing-default, fig.height=4.2} if (requireNamespace("fect", quietly = TRUE)) { d <- subset(sim_base, id == 102) d$Y[d$time %in% c(10, 11, 12)] <- NA d$Y[d$time %in% c(30, 31)] <- NA panelview(Y ~ D, data = d, index = c("id", "time"), type = "outcome", main = "leave.gap = FALSE: dotted bridge across missing periods") } ``` ### Interior gap: clean break under `leave.gap = TRUE` Same data, but `leave.gap = TRUE` keeps the NA rows, so the line breaks at every gap with no bridge. ```{r outcome-missing-leave-gap, fig.height=4.2} if (requireNamespace("fect", quietly = TRUE)) { d <- subset(sim_base, id == 102) d$Y[d$time %in% c(10, 11, 12)] <- NA d$Y[d$time %in% c(30, 31)] <- NA panelview(Y ~ D, data = d, index = c("id", "time"), type = "outcome", leave.gap = TRUE, main = "leave.gap = TRUE: line breaks at gaps, no bridge") } ``` ### Missing at the boundary Drop unit 102's $Y$ at $t=1\!-\!3$ (the start of its series) and pass `leave.gap = TRUE` so the row is retained, the x-axis stays at $t=1\!-\!35$, and the topology classifier still sees the early treated $D$ values. The line is empty at $t=1\!-\!3$; everything else is unchanged. ```{r outcome-missing-boundary, fig.height=4.2} if (requireNamespace("fect", quietly = TRUE)) { d <- subset(sim_base, id == 102) d$Y[d$time %in% 1:3] <- NA panelview(Y ~ D, data = d, index = c("id", "time"), type = "outcome", leave.gap = TRUE, main = "Drop Y at t=1-3 (leave.gap = TRUE keeps x-axis and classification)") } ``` Under the default `leave.gap = FALSE`, the same NA injection drops those rows entirely. The panel for this unit shrinks to $t=4\!-\!35$, its surviving $D$ pattern becomes monotonic, and the classifier re-types it as staggered. Use `leave.gap = TRUE` whenever you need the boundary intact. ### `by.group = TRUE` with all three subpanels `by.group = TRUE` splits units by their treatment history: always under control, always under treatment, and treatment status changed. To exercise all three subpanels we promote one `sim_base` unit to always-treated. ```{r outcome-bygroup-full, fig.height=8} if (requireNamespace("fect", quietly = TRUE)) { d <- sim_base d$D[d$id == 101] <- 1 ## promote unit 101 to always-treated panelview(Y ~ D, data = d, index = c("id", "time"), type = "outcome", by.group = TRUE, main = "by.group = TRUE: always control / always treated / status changed") } ``` ## Discrete outcomes Set `outcome.type = "discrete"` when the outcome takes a small number of integer values. ```{r outcome3-1, fig.height=5} panelview(Y ~ D, data = simdata, index = c("id", "time"), outcome.type = "discrete", type = "outcome", xlim = c(8, 15)) ``` Split by treatment-status group. ```{r outcome3-2, fig.height=5} panelview(Y ~ D, data = simdata, index = c("id", "time"), by.group = TRUE, outcome.type = "discrete", type = "outcome", xlim = c(8, 15)) ``` ## Multi-level or continuous treatment When the treatment variable has more than 5 unique values (e.g., `polity2`), treatment status is not shown on the outcome plot. ```{r outcome3-3, fig.height=5} panelview(Capacity ~ polity2 + lngdp, data = capacity, index = c("ccode", "year"), main = "Measuring State Capacity", type = "outcome", legendOff = TRUE) ```