I am trying to test stationarity for a panel of 15 countries current account (1980–2012) annual data in Stata. I have chosen the LM unit root test by Lluís Carrion-i-Silvestre et al. (2005; to consider structural breaks in the panels), and the non-linear ADF unit root test based on their statistical power.

Trouble is, I am failing to figure out how to conduct these tests in Stata. Can anyone help please with suggestions on how to run the tests in Stata please?

  • 2
    $\begingroup$ Please post a complete reference, as well as a reference dataset. $\endgroup$ – tchakravarty May 8 '14 at 11:12
  • $\begingroup$ @ fg nu, Lluís Carrion-i-Silvestre, J., et al. (2005). "Breaking the panels: An application to the GDP per capita." Econometrics Journal 8(2): 159-175. and Kapetanios, G., et al. (2003). "Testing for a unit root in the nonlinear STAR framework." Journal of Econometrics 112(2): 359-379 $\endgroup$ – sunga09 May 9 '14 at 0:14

I do not believe these tests are currently available. See help xtunitroot for details about the available Levin-Lin-Chu, Harris-Tzavalis, Breitung, Im-Pesaran-Shin, Fisher-type, and Hadri Lagrange multiplier tests.

A search for Stata-linked user contributions via findit Lluís Carrion-i-Silvestre, findit Carrion, or findit Silvestre returns no results, so unless the test goes by some other name or reference than the first author of this paper, it's unlikely to have been implemented for Stata as a package.

Should you desire an answer to provide code within Stata's programming environments, you should provide enough details of the test for respondents unfamiliar with the test to provide an approach.

  • 3
    $\begingroup$ It such tests really aren't available, this would be an answer. You may want to simply expand this a little. $\endgroup$ – gung - Reinstate Monica May 8 '14 at 15:18
  • $\begingroup$ Lack of availability, which in the meantime is useful information, indicates that a good answer will either provide enough information to code the algorithm in one of Stata's programming languages or else it will point to existing non-Stata code that could be used in conjunction with Stata. $\endgroup$ – whuber May 8 '14 at 15:23
  • $\begingroup$ @ gung @ whuber and @Alexis the LM test in question runs the regression x_it=α_i+∑_(k=1)^(m_i)▒〖〖γ_(i,k) DU〗_(i,k,t)+β_i t+∑_(k=1)^(m_i)▒〖〖θ_(i,k) DT〗_(i,k,t)^*+ε_(i,t) 〗〗 where DU and DT are intercept and slope dummies repsectively. The NLADF runs the auxilliary regression ∆x_(i,t)=δ_i x_(i,t-1)^3+∑_(j=1)^(p_i)▒〖ρ_(i,j) 〖∆x〗_(i,t-j)+ε_(i,t) 〗and test for H_0: δ=0. However the original regression which has untestable unit root is ∆x_t=〖∅x〗_(t-1)+〖∂x〗_(t-1) [1-exp⁡(〖-θx〗_(t-d)^2)]+∑_(i=1)^k▒〖δ_k ∆x_(t-i)+ε_t 〗. Thanks in advance for your help and interest in my question. $\endgroup$ – sunga09 May 9 '14 at 0:25
  • 1
    $\begingroup$ OMFG. Where did you cut and paste it from??? Post all this formula stuff in the original question. Nobody will even try to read it in the comment and in such ugly formatting. $\endgroup$ – StasK Sep 6 '14 at 22:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.