robustreg

Last updated on 4th February, 2026.

robustreg is useful for specification curve analysis, meta analysis, and robustness. It systematically runs a series of regressions based on user-defined specification combinations, plots the results, and saves the output to a dataset.

An example of the final curve :

---

Installation

robustreg is hosted on GitHub. You can install it directly in Stata by running the following command:

net install robustreg, from("https://raw.githubusercontent.com/maobin-xu/robustreg/main/") replace

---

Syntax

The basic syntax for robustreg is:

robustreg, dep(str) indep(str) [options]

---

Options

Regression Options (using reghdfe)

• dep(str): A list of dependent variables.
• indep(str): A list of sets of independent variables. Minimum number of sets is 1. Allow multiple variables within each set. If you use multiple sets, specification curve will only use the estimated coefficients for the first variable within each set
• sample(str): A list of sample variables (that equal 1) to define the samples used in regressions.
• control(str): A list of control variable sets.
• fe(str): A list of fixed effects.
• se(str): A list of standard error clustering methods.
• reghdfe_opt(str): Any other valid options for the reghdfe command.

Plot Options (using twoway and graph export)

• plot(str): Path to save the regression specification curve graph. This option is required to generate the plot.
• order(str): The order of specifications for the curve. The default is order(dep, indep, control, fe, se, sample).
• orderl(str): A list of labels for the order of specifications.
• depl(str): A list of labels for the dependent variables.
• indepl(str): A list of labels for the key independent variables.
• samplel(str): A list of labels for the sample options.
• controll(str): A list of labels for the control variable sets.
• fel(str): A list of labels for the fixed effects.
• sel(str): A list of labels for the standard error clustering methods.
• level(#): The confidence level for the confidence intervals. Default is 95.
• twoway_opt(str): Other options for the twoway command.
• scatter_opt(str): Other options for the twoway scatter command, used to plot the β coefficients.
• rarea_opt(str): Other options for the twoway rarea command, used to plot the confidence intervals.
• graph_opt(str): Other options for the graph export command.

Save Options

• save(str): Path to save the regression results as a Stata dataset (.dta).
• stats(str): A list of additional e() statistics from reghdfe to save.

---

Examples

First, let's generate some sample data to work with.

clear
set obs 1000

// Key independent variables
gen x1 = rnormal()
gen x2 = x1 + 0.2*rnormal()

// Control variables
gen c1 = rnormal()*2
gen c2 = runiform()-0.5 

// Dependent variables
gen e = rnormal()/5
gen y1 = 0.24*x1 + 0.4*exp(c1) + e
gen y2 = 0.26*x1 + 0.4*exp(c1) + e

// Sample variables
gen s1 = mod(_n, 2) == 1
gen s2 = mod(_n-1, 3) + 1 == 1

// Fixed effect variables
gen f1 = mod(_n, 50)
gen f2 = mod(_n, 100)

Example 1: Save Regression Results Only

This command runs all possible regressions and saves the results to regtab1.dta, without generating a plot.

robustreg , dep(y1, y2) indep(x1, x1 x2) control(c1, , c1 c2) fe( , f1, f2) sample( , s1) se(robust, , f2) save("regtab1")

The results are as follows:

Example 2: Generate and Save a Plot

This command generates a specification curve and saves it as curve1.png.

robustreg , dep(y1, y2) indep(x1) control(c1, , c1 c2) fe( , f1, f2) sample( , s1)  plot("curve1.png") twoway_opt(graphregion(margin(l=42 r=5 t=0 b=0))) graph_opt(width(1500) height(1500))

Example 3: Full Customization

This example shows the command with a full set of customized options, including specific labels for each option set.

robustreg, ///
    dep(y1, y2) indepl("Y1" "Y2") ///
    indep(x1, x2) indepl("X1" "X2") ///
    control( , c1, c1 c2) controll("No" "Control set 1" "Control set 2") ///
    fe( , f1, f2) fel("No" "Fixed effects 1" "Fixed effects 2") ///
    sample( , s1) samplel("Full sample" "Sample 1") ///
    se( , robust, f2) sel("No" "Robust" "F2") ///
    save("regtab2.dta") plot("curve2.png") level(90) ratio(1) fontsize(small) ///
    twoway_opt(graphregion(margin(l=22 r=5 t=0 b=0)) scale(0.6)) ///
    rarea_opt(color(gray%40)) scatter_opt(color(gray)) ///
    graph_opt(width(3000) height(2000))