Time series and forecasting

Start with the expanded example for each command. Open the variants for options and alternative R approaches. Read the qualifications before adapting the code.

ac

Autocorrelation function

Stata

ac y, lags(12)

R stats

stats::acf(y, lag.max = 12, plot = FALSE)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. R returns numeric correlations; request plotting separately.

Stata reference · R reference

arch

GARCH(1,1) conditional-variance model

Stata

arch y, arch(1) garch(1) noconstant

R tseries

tseries::garch(as.numeric(y), order = c(1, 1), trace = FALSE)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Fit a mean model separately; optimization and initialization differ.

Stata reference · R reference

ARCH(1) conditional-variance model

Stata

arch y, arch(1) noconstant

R tseries

tseries::garch(as.numeric(y), order = c(0, 1), trace = FALSE)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. tseries::garch models zero-mean observations; remove or explicitly model the conditional mean first.

Stata reference · R reference

arfima

Fractionally integrated ARMA estimation

Stata

arfima y, ar(1) ma(1)

R fracdiff

fracdiff::fracdiff(y, nar = 1, nma = 1)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. fracdiff uses approximate likelihood and a different MA sign convention.

Stata reference · R reference

arfimasoc

Fractional-ARMA order comparison

Stata

arfimasoc y, maxar(1) maxma(1)

R fracdiff

{ grid <- expand.grid(p = 0:1, q = 0:1); grid$AIC <- mapply(function(p, q) { f <- fracdiff::fracdiff(y, nar = p, nma = q); -2 * f$log.likelihood + 2 * (p + q + 2) }, grid$p, grid$q); grid }

Fixed workflow; adapt the named data columns explicitly. Approximate-likelihood order-selection workflow; no numerical parity claim.

Stata reference · R reference

arima

ARMAX with external regressor

Stata

arima y x, arima(1,0,0)

R stats

stats::arima(y, order = c(1, 0, 0), xreg = x)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

11 more examples and alternatives

Seasonally differenced moving average

Stata

arima y, arima(0,0,0) sarima(0,1,1,4)

R stats

stats::arima(y, order = c(0, 0, 0), seasonal = list(order = c(0, 1, 1), period = 4))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

Seasonal autoregression

Stata

arima y, arima(0,0,0) sarima(1,0,0,4)

R stats

stats::arima(y, order = c(0, 0, 0), seasonal = list(order = c(1, 0, 0), period = 4))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

MA(1) with intercept

Stata

arima y, arima(0,0,1)

R stats

stats::arima(y, order = c(0, 0, 1))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

Integrated moving-average model

Stata

arima y, arima(0,1,1)

R stats

stats::arima(y, order = c(0, 1, 1))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

AR(1) with intercept

Stata

arima y, arima(1,0,0)

R stats

stats::arima(y, order = c(1, 0, 0))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

State-space maximum-likelihood estimation

Stata

arima y, arima(1,0,0) diffuse

R stats

stats::arima(y, order = c(1, 0, 0), method = "ML", SSinit = "Rossignol2011")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

Zero-mean autoregression

Stata

arima y, arima(1,0,0) noconstant

R stats

stats::arima(y, order = c(1, 0, 0), include.mean = FALSE)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

ARMA(1,1)

Stata

arima y, arima(1,0,1)

R stats

stats::arima(y, order = c(1, 0, 1))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

ARMA estimation using ML

Stata

arima y, arima(1,0,1)

R stats

stats::arima(y, order = c(1, 0, 1), method = "ML")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Explicit likelihood setting avoids relying on R’s default CSS-ML initialization.

Stata reference · R reference

ARMA estimation using CSS

Stata

arima y, arima(1,0,1) // conditional least-squares alternative

R stats

stats::arima(y, order = c(1, 0, 1), method = "CSS")

Fixed workflow; adapt the named data columns explicitly. Conditional sum of squares is a workflow alternative, not Stata’s default likelihood.

Stata reference · R reference

ARIMA with first difference

Stata

arima y, arima(1,1,0)

R stats

stats::arima(y, order = c(1, 1, 0))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Parameterization, initialization and likelihood conventions differ; compare selected model, not default output.

Stata reference · R reference

arimasoc

ARMA information-criterion grid

Stata

arimasoc y, maxar(2) maxma(2)

R stats

{ grid <- expand.grid(p = 0:2, q = 0:2); grid$AIC <- mapply(function(p, q) stats::AIC(stats::arima(y, order = c(p, 0, q), method = "ML")), grid$p, grid$q); grid }

Fixed workflow; adapt the named data columns explicitly. Comparable-sample model-selection workflow; R and Stata likelihood constants may differ.

Stata reference · R reference

corrgram

Autocorrelation and partial-autocorrelation table

Stata

corrgram y, lags(8)

R stats

list(ac = stats::acf(y, lag.max = 8, plot = FALSE), pac = stats::pacf(y, lag.max = 8, plot = FALSE))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. R returns numeric correlations; request plotting separately.

Stata reference · R reference

cumsp

Cumulative spectral distribution

Stata

cumsp y

R stats

{ sp <- stats::spectrum(y, plot = FALSE, taper = 0); data.frame(frequency = sp$freq, cumulative = cumsum(sp$spec) / sum(sp$spec)) }

Fixed workflow; adapt the named data columns explicitly. Computes curve data; confidence bands are not included.

Stata reference · R reference

dfgls

DF-GLS unit-root test with trend

Stata

dfgls y, maxlag(4)

R urca

urca::ur.ers(y, type = "DF-GLS", model = "trend", lag.max = 4)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Lag-selection and finite-sample critical values may differ.

Stata reference · R reference

DF-GLS unit-root test with constant

Stata

dfgls y, maxlag(4) notrend

R urca

urca::ur.ers(y, type = "DF-GLS", model = "constant", lag.max = 4)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Lag-selection and finite-sample critical values may differ.

Stata reference · R reference

dfuller

ADF unit-root test with drift and 0 augmentation lags

Stata

dfuller y, lags(0)

R urca

urca::ur.df(y, type = "drift", lags = 0)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Match deterministic terms, sample window and lag selection; critical-value implementations can differ.

Stata reference · R reference

ADF unit-root test with none and 0 augmentation lags

Stata

dfuller y, lags(0) noconstant

R urca

urca::ur.df(y, type = "none", lags = 0)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Match deterministic terms, sample window and lag selection; critical-value implementations can differ.

Stata reference · R reference

ADF unit-root test with trend and 0 augmentation lags

Stata

dfuller y, lags(0) trend

R urca

urca::ur.df(y, type = "trend", lags = 0)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Match deterministic terms, sample window and lag selection; critical-value implementations can differ.

Stata reference · R reference

ADF unit-root test with drift and 2 augmentation lags

Stata

dfuller y, lags(2)

R urca

urca::ur.df(y, type = "drift", lags = 2)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Match deterministic terms, sample window and lag selection; critical-value implementations can differ.

Stata reference · R reference

ADF unit-root test with none and 2 augmentation lags

Stata

dfuller y, lags(2) noconstant

R urca

urca::ur.df(y, type = "none", lags = 2)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Match deterministic terms, sample window and lag selection; critical-value implementations can differ.

Stata reference · R reference

1 more examples and alternatives

ADF unit-root test with trend and 2 augmentation lags

Stata

dfuller y, lags(2) trend

R urca

urca::ur.df(y, type = "trend", lags = 2)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Match deterministic terms, sample window and lag selection; critical-value implementations can differ.

Stata reference · R reference

fcast

Iterated bivariate VAR point forecasts

Stata

fcast compute pred, step(8)

R stats

{ out <- matrix(NA_real_, 8, 2); last <- tail(Y, 1); intercept <- stats::coef(fit)[1, ]; for (h in 1:8) { last <- matrix(intercept + drop(A %*% drop(last)), nrow = 1); out[h, ] <- last }; colnames(out) <- colnames(Y); out }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Point forecasts only; no system forecast-error covariance or interval calculation.

Stata reference · R reference

forecast

Eight-period ARIMA forecast

Stata

forecast solve, begin(161)

R forecast

forecast::forecast(fit, h = 8)

Fixed workflow; adapt the named data columns explicitly. Single-equation forecasting workflow; forecast system identities are not translated.

Stata reference · R reference

Forecast with ninety-percent prediction intervals

Stata

forecast solve, begin(161) periods(8) simulate(errors, statistic(stddev, prefix(sd_)))

R forecast

forecast::forecast(fit, h = 8, level = 90)

Fixed workflow; adapt the named data columns explicitly. Single-equation forecasting workflow; forecast system identities are not translated. R produces analytic 90% prediction intervals; the Stata example instead simulates forecast standard deviations, requiring a separate interval calculation.

Stata reference · R reference

One-step fitted values from ARIMA

Stata

forecast solve, begin(161) static

R forecast

stats::fitted(forecast::Arima(y, order = c(1, 0, 0)))

Fixed workflow; adapt the named data columns explicitly. In-sample one-step prediction workflow; Stata’s forecast model setup is separate.

Stata reference · R reference

irf

Reduced-form VAR(1) impulse-response matrices

Stata

irf create model, step(8)

R base

{ powers <- vector("list", 9); powers[[1]] <- diag(2); for (h in 2:9) powers[[h]] <- powers[[h - 1]] %*% A; powers }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Unit reduced-form innovation responses; orthogonalized structural responses require an identification scheme.

Stata reference · R reference

Cholesky-orthogonalized VAR responses

Stata

irf create model, step(8) // orthogonalized responses

R base

{ impact <- t(chol(crossprod(stats::residuals(fit)) / stats::nobs(fit))); powers <- vector("list", 9); powers[[1]] <- impact; for (h in 2:9) powers[[h]] <- A %*% powers[[h - 1]]; powers }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Recursive ordering is y then x; covariance divisor and identifying assumptions matter.

Stata reference · R reference

ivlpirf

Instrumental-variable local-projection point responses

Stata

ivlpirf y, endogenous(x = z) step(8) lags(1)

R AER

vapply(0:8, function(h) { i <- 2:(nrow(data) - h); d <- data.frame(outcome = data$y[i + h], impulse = data$x[i], Ly = data$y[i - 1], instrument = data$z[i]); unname(stats::coef(AER::ivreg(outcome ~ impulse + Ly | instrument + Ly, data = d))[2]) }, numeric(1))

Fixed workflow; adapt the named data columns explicitly. Horizonwise 2SLS workflow; instrument validity, normalization and HAC inference require separate decisions.

Stata reference · R reference

lpirf

Local-projection dynamic response to an external impulse

Stata

lpirf y, exog(x) step(8) lags(1)

R stats

vapply(0:8, function(h) { i <- 2:(nrow(data) - h); unname(stats::coef(stats::lm(data$y[i + h] ~ data$x[i] + data$y[i - 1] + data$x[i - 1]))[2]) }, numeric(1))

Fixed workflow; adapt the named data columns explicitly. Simple horizonwise point-estimation workflow; HAC inference and system restrictions must be added for research use.

Stata reference · R reference

newey

Newey-West HAC coefficient inference

Stata

newey y x, lag(4)

R sandwich

{ fit <- stats::lm(y ~ x, data = data); lmtest::coeftest(fit, vcov. = sandwich::NeweyWest(fit, lag = 4, prewhite = FALSE, adjust = TRUE)) }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Explicit bandwidth, prewhitening and finite-sample correction; align those choices.

Stata reference · R reference

pac

Partial autocorrelation function

Stata

pac y, lags(12)

R stats

stats::pacf(y, lag.max = 12, plot = FALSE)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. R returns numeric correlations; request plotting separately.

Stata reference · R reference

pergram

Raw periodogram

Stata

pergram y

R stats

stats::spectrum(y, plot = FALSE, taper = 0, detrend = FALSE)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Normalization and default tapering can differ.

Stata reference · R reference

pperron

Phillips-Perron Z-alpha test with constant

Stata

pperron y, lags(4)

R urca

urca::ur.pp(y, type = "Z-alpha", model = "constant", use.lag = 4)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Stata reports several statistics together; R selects one statistic. Kernel and deterministic terms must be matched.

Stata reference · R reference

Phillips-Perron Z-tau test with constant

Stata

pperron y, lags(4)

R urca

urca::ur.pp(y, type = "Z-tau", model = "constant", use.lag = 4)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Stata reports several statistics together; R selects one statistic. Kernel and deterministic terms must be matched.

Stata reference · R reference

Phillips-Perron Z-alpha test with trend

Stata

pperron y, lags(4) trend

R urca

urca::ur.pp(y, type = "Z-alpha", model = "trend", use.lag = 4)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Stata reports several statistics together; R selects one statistic. Kernel and deterministic terms must be matched.

Stata reference · R reference

Phillips-Perron Z-tau test with trend

Stata

pperron y, lags(4) trend

R urca

urca::ur.pp(y, type = "Z-tau", model = "trend", use.lag = 4)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Stata reports several statistics together; R selects one statistic. Kernel and deterministic terms must be matched.

Stata reference · R reference

predict

Dynamic ARIMA forecast and standard errors

Stata

predict yhat, y dynamic(161)

R stats

stats::predict(fit, n.ahead = 8)

Fixed workflow; adapt the named data columns explicitly. Single-equation forecasting workflow; forecast system identities are not translated.

Stata reference · R reference

psdensity

Parametric AR(1) spectral density

Stata

arima y, arima(1,0,0)
psdensity density, n(64)

R stats

{ phi <- unname(fit$coef["ar1"]); omega <- seq(0, pi, length.out = 64); data.frame(frequency = omega / (2 * pi), density = fit$sigma2 / (2 * pi * Mod(1 - phi * exp(-1i * omega))^2)) }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Frequency and density normalization conventions must be reconciled.

Stata reference · R reference

rolling

Expanding-window recursive regression

Stata

rolling _b, window(40) recursive: regress y x

R stats

t(vapply(40:nrow(data), function(i) stats::coef(stats::lm(y ~ x, data[seq_len(i), ])), numeric(2)))

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Uses a fixed first observation and increasing estimation endpoint.

Stata reference · R reference

Rolling-window regression coefficients

Stata

rolling _b, window(40): regress y x

R zoo

zoo::rollapply(data, width = 40, FUN = function(z) stats::coef(stats::lm(y ~ x, data = as.data.frame(z))), by.column = FALSE, align = "right")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Output rows correspond to complete right-aligned windows; failures need explicit handling in production.

Stata reference · R reference

Rolling sample volatility

Stata

rolling sd=r(sd), window(20): summarize y

R zoo

zoo::rollapply(y, width = 20, FUN = stats::sd, align = "right")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Sample standard deviation, not model-implied conditional volatility.

Stata reference · R reference

sspace

Local-linear-trend state-space model

Stata

sspace (level L.level L.slope, state noconstant) (slope L.slope, state noconstant) (y level, noconstant)

R stats

stats::StructTS(y, type = "trend")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Specific local-linear-trend model; initialization can differ.

Stata reference · R reference

Local-level state-space model

Stata

sspace (level L.level, state noconstant) (y level, noconstant)

R stats

stats::StructTS(y, type = "level")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Specific local-level model equivalent; sspace supports a much wider model class.

Stata reference · R reference

tsfilter

Baxter-King approximate band-pass cycle

Stata

tsfilter bk cycle = y, minperiod(6) maxperiod(32) smaorder(12)

R stats

{ K <- 12; j <- -K:K; low <- 2 * pi / 32; high <- 2 * pi / 6; w <- ifelse(j == 0, (high - low) / pi, (sin(high * j) - sin(low * j)) / (pi * j)); w <- w - mean(w); stats::filter(y, w, sides = 2) }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Symmetric finite-order band-pass filter loses endpoint observations.

Stata reference · R reference

Hodrick-Prescott trend and cycle decomposition

Stata

tsfilter hp cycle = y, smooth(1600) trend(trend)

R base

{ n <- length(y); D <- diff(diag(n), differences = 2); trend <- drop(solve(diag(n) + 1600 * crossprod(D), as.numeric(y))); data.frame(trend = trend, cycle = as.numeric(y) - trend) }

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Complete regular series; this direct penalized least-squares solution uses the two-sided HP filter.

Stata reference · R reference

tsreport

Locate missing time periods

Stata

tsreport, report

R base

data.frame(previous = head(data$t, -1), next_time = tail(data$t, -1), gap = diff(data$t))[diff(data$t) != 1, ]

Fixed workflow; adapt the named data columns explicitly. Assumes integer unit-spaced time; date calendars need the intended interval definition.

Stata reference · R reference

tssmooth

Simple exponential smoothing

Stata

tssmooth exponential smooth = y, parms(.3)

R stats

stats::HoltWinters(y, alpha = .3, beta = FALSE, gamma = FALSE)

Fixed workflow; adapt the named data columns explicitly. Initial values, endpoints and trend formulations differ; align before comparing.

Stata reference · R reference

Holt trend smoothing workflow

Stata

tssmooth hwinters smooth = y

R stats

stats::HoltWinters(y, gamma = FALSE)

Fixed workflow; adapt the named data columns explicitly. Initial values, endpoints and trend formulations differ; align before comparing.

Stata reference · R reference

Centered five-term moving average

Stata

tssmooth ma smooth = y, window(2 1 2)

R stats

stats::filter(y, rep(1 / 5, 5), sides = 2)

Fixed workflow; adapt the named data columns explicitly. Initial values, endpoints and trend formulations differ; align before comparing.

Stata reference · R reference

Trailing four-term moving average

Stata

tssmooth ma smooth = y, window(3 1 0)

R stats

stats::filter(y, rep(1 / 4, 4), sides = 1)

Fixed workflow; adapt the named data columns explicitly. Initial values, endpoints and trend formulations differ; align before comparing.

Stata reference · R reference

Additive seasonal Holt-Winters smoothing

Stata

tssmooth shwinters smooth = y

R stats

stats::HoltWinters(y, seasonal = "additive")

Fixed workflow; adapt the named data columns explicitly. Initial values, endpoints and trend formulations differ; align before comparing.

Stata reference · R reference

1 more examples and alternatives

Running-median nonlinear smoother

Stata

tssmooth nl smooth = y, smoother(3)

R stats

stats::runmed(as.numeric(y), k = 3, endrule = "median")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Endpoint rules should be checked separately before comparing.

Stata reference · R reference

ucm

Unobserved components: level

Stata

ucm y, model(llevel)

R stats

stats::StructTS(y, type = "level")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. R supports selected structural models; Stata ucm permits additional components and constraints.

Stata reference · R reference

Unobserved components: trend

Stata

ucm y, model(lltrend)

R stats

stats::StructTS(y, type = "trend")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. R supports selected structural models; Stata ucm permits additional components and constraints.

Stata reference · R reference

Unobserved components: BSM

Stata

ucm y, model(lltrend) seasonal(4)

R stats

stats::StructTS(y, type = "BSM")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. R supports selected structural models; Stata ucm permits additional components and constraints.

Stata reference · R reference

var

Unrestricted bivariate VAR(1) by equation OLS

Stata

var y x, lags(1)

R stats

stats::lm(cbind(y, x) ~ Ly + Lx, data = lagged)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Coefficient equivalent for an unrestricted common-sample VAR; system covariance reporting differs.

Stata reference · R reference

VAR with deterministic time trend

Stata

var y x, lags(1) exog(trend)

R stats

stats::lm(cbind(y, x) ~ Ly + Lx + seq_len(nrow(lagged)), data = lagged)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Trend origin affects intercepts; compare the same sample and trend indexing.

Stata reference · R reference

Zero-intercept bivariate VAR

Stata

var y x, lags(1) noconstant

R stats

stats::lm(cbind(y, x) ~ 0 + Ly + Lx, data = lagged)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Specify identical lag matrices and deterministic terms.

Stata reference · R reference

vargranger

Test whether y Granger-predicts x

Stata

var y x, lags(1)
vargranger

R lmtest

lmtest::grangertest(x ~ y, order = 1, data = data)

Differs from Stata Fit var y x, lags(1) before vargranger; Stata VAR otherwise defaults to lags(1 2). grangertest uses an F reference distribution, whereas vargranger reports a Wald chi-square; match direction and sample, and do not interpret predictability as structural causation.

Stata reference · R reference

Test whether x Granger-predicts y

Stata

var y x, lags(1)
vargranger

R lmtest

lmtest::grangertest(y ~ x, order = 1, data = data)

Differs from Stata Fit var y x, lags(1) before vargranger; Stata VAR otherwise defaults to lags(1 2). grangertest uses an F reference distribution, whereas vargranger reports a Wald chi-square; match direction and sample, and do not interpret predictability as structural causation.

Stata reference · R reference

varstable

Eigenvalues of the VAR(1) companion matrix

Stata

varstable

R base

eigen(A, only.values = TRUE)$values

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Stability requires all companion roots strictly inside the unit circle.

Stata reference · R reference

vec

Rank-one vector error-correction model

Stata

vec y x, rank(1) lags(2)

R urca

urca::cajorls(urca::ca.jo(Y, type = "trace", ecdet = "none", K = 2, spec = "transitory"), r = 1)

Differs from Stata urca ecdet=“const” matches Stata trend(rconstant); ecdet=“none” includes an unrestricted intercept and matches trend(constant); ecdet=“trend” matches trend(rtrend). ca.jo does not provide Stata trend(none) or trend(trend). Match cointegrating-vector normalization before comparing coefficients.

Stata reference · R reference

vecrank

Test cointegration rank with an unrestricted intercept

Stata

vecrank y x, lags(2) trend(constant)

R urca

urca::ca.jo(Y, type = "trace", ecdet = "none", K = 2, spec = "transitory")

Differs from Stata urca ecdet=“const” matches Stata trend(rconstant); ecdet=“none” includes an unrestricted intercept and matches trend(constant); ecdet=“trend” matches trend(rtrend). ca.jo does not provide Stata trend(none) or trend(trend). Match cointegrating-vector normalization before comparing coefficients.

Stata reference · R reference

Johansen trace rank test with a restricted constant

Stata

vecrank y x, lags(2) trend(rconstant)

R urca

urca::ca.jo(Y, type = "trace", ecdet = "const", K = 2, spec = "transitory")

Differs from Stata urca ecdet=“const” matches Stata trend(rconstant); ecdet=“none” includes an unrestricted intercept and matches trend(constant); ecdet=“trend” matches trend(rtrend). ca.jo does not provide Stata trend(none) or trend(trend). Match cointegrating-vector normalization before comparing coefficients.

Stata reference · R reference

Johansen maximum-eigenvalue rank test with a restricted constant

Stata

vecrank y x, lags(2) trend(rconstant) max

R urca

urca::ca.jo(Y, type = "eigen", ecdet = "const", K = 2, spec = "transitory")

Differs from Stata urca ecdet=“const” matches Stata trend(rconstant); ecdet=“none” includes an unrestricted intercept and matches trend(constant); ecdet=“trend” matches trend(rtrend). ca.jo does not provide Stata trend(none) or trend(trend). Match cointegrating-vector normalization before comparing coefficients.

Stata reference · R reference

Johansen trace rank test with a restricted trend

Stata

vecrank y x, lags(2) trend(rtrend)

R urca

urca::ca.jo(Y, type = "trace", ecdet = "trend", K = 2, spec = "transitory")

Differs from Stata urca ecdet=“const” matches Stata trend(rconstant); ecdet=“none” includes an unrestricted intercept and matches trend(constant); ecdet=“trend” matches trend(rtrend). ca.jo does not provide Stata trend(none) or trend(trend). Match cointegrating-vector normalization before comparing coefficients.

Stata reference · R reference

Johansen maximum-eigenvalue rank test with a restricted trend

Stata

vecrank y x, lags(2) trend(rtrend) max

R urca

urca::ca.jo(Y, type = "eigen", ecdet = "trend", K = 2, spec = "transitory")

Differs from Stata urca ecdet=“const” matches Stata trend(rconstant); ecdet=“none” includes an unrestricted intercept and matches trend(constant); ecdet=“trend” matches trend(rtrend). ca.jo does not provide Stata trend(none) or trend(trend). Match cointegrating-vector normalization before comparing coefficients.

Stata reference · R reference

1 more examples and alternatives

Test cointegration rank with an unrestricted intercept

Stata

vecrank y x, lags(2) trend(constant) max

R urca

urca::ca.jo(Y, type = "eigen", ecdet = "none", K = 2, spec = "transitory")

Differs from Stata urca ecdet=“const” matches Stata trend(rconstant); ecdet=“none” includes an unrestricted intercept and matches trend(constant); ecdet=“trend” matches trend(rtrend). ca.jo does not provide Stata trend(none) or trend(trend). Match cointegrating-vector normalization before comparing coefficients.

Stata reference · R reference

wntestq

Ljung-Box test after AR(1) estimation

Stata

predict e, residuals
wntestq e, lags(8)

R stats

stats::Box.test(stats::residuals(fit), lag = 8, type = "Ljung-Box", fitdf = 1)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. R explicitly deducts the estimated AR parameter from the reference degrees of freedom.

Stata reference · R reference

Ljung-Box serial-correlation portmanteau test

Stata

wntestq y, lags(8)

R stats

stats::Box.test(y, lag = 8, type = "Ljung-Box")

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Use fitdf for tests on estimated-model residuals.

Stata reference · R reference

xcorr

Cross-correlation at positive and negative lags

Stata

xcorr y x, lags(8)

R stats

stats::ccf(y, x, lag.max = 8, plot = FALSE)

Differs from Stata Fixed workflow; adapt the named data columns explicitly. Check lag sign conventions before interpreting lead-lag direction.

Stata reference · R reference