LSUnit (LM unit root test with multiple breaks)

[TomDoan]

This implements a number of varieties of LM unit root tests allowing for 0, 1, 2 or even more structural breaks of one of two varieties, as described in Lee and Strazicich(2003), "Minimum LM Unit Root Test with Two Structural Breaks," Review of Economics and Statistics, vol 85, no. 4, pp 1082-1089. This differs slightly from Lee and Strazicich's Gauss code as the lag pruning regressions use a common estimation range (from maxlag+1 on) rather than the lag+1 samples used in the Gauss code. The method chosen here would make it simpler to change the pruning method to an alternative like minimum AIC.

I incorporated a number of suggestions made by Junsoo Lee regarding the output, which includes output of all the coefficients other than the auxiliary lags.

Note that this can take a very long time for two or more breaks if the data set is large (>500 observations). The calculation time goes up with T^(# of breaks+1), so with 1000 data points, you're looking at roughly 1000 times as long to do two breaks as one.

Tom Doan
Estima

Procedure file:

lsunit.src

(9.67 KiB) Downloaded 659 times

Example:

Code: Select all

source lsunit.src
data(format=free,unit=input,org=cols) 1 38 year x
1960    11508
1961    11369
1962    11930
1963    12636
1964    13194
1965    13264
1966    13875
1967    14193
1968    15172
1969    15712
1970    16316
1971    16872
1972    17563
1973    18489
1974    18606
1975    18741
1976    18728
1977    18480
1978    19303
1979    19682
1980    19968
1981    20271
1982    19464
1983    20127
1984    20712
1985    21438
1986    21719
1987    22443
1988    23040
1989    22995
1990    22226
1991    21921
1992    22367
1993    23176
1994    24144
1995    24818
1996    25077
1997    25608
set logx = log(x)
@lsunit(breaks=2,pi=.10,lags=5,method=gtos,model=crash) logx
@lsunit(breaks=2,pi=.10,lags=5,method=gtos,model=break) logx

[Jose]

Hello TomDoan!

I'm using the procedure of rats to "Lee-Strazicich Unit Root" and I have a doubt about the output. An Example:

Lee-Strazicich Unit Root Test, Series QCVEN
Regression Run From 1971:03 to 2006:04
Observations 141
Crash Model with 1 breaks
With 1 chosen from 5

Variable Coefficient Std Error
S{1} -0.0896 -2.7737
Constant 0.0062 0.9871
D(1986:04) 0.2750 4.2558


My doubt is whether the numbers in the column "Std Error" are standard errors rather than t-student statistics. This numbers are really the standards errors?

Thanks for your attention,

Anchieta
Obs: I have post this same message in other topic, ok?

[TomDoan]

You're correct. Those are the t-statistics. I've corrected the code shown in the previous message on 28 April 2008.

[ecsaa]

I am relatively new to RATS programming.

Can you tell me whether this program code can be adapted to report:

1) the coefficient and t-ratio of the trend term;

2) the number of lags in the estimated equation.

Many thanks for any assistance.

[TomDoan]

Code: Select all

Lee-Strazicich Unit Root Test, Series LOGX
Regression Run From 7 to 38
Observations          32
Trend Break Model with 2 breaks
With 2 chosen from 5

Variable     Coefficient T-Stat
S{1}             -1.0983 -5.1986
Constant          0.0198  2.4096
D(11)            -0.0152 -0.8670
DT(11)            0.0272  2.0811
D(31)            -0.0146 -0.7537
DT(31)           -0.0119 -1.3742

The "With 2 chosen from 5" is missing the word "lags", but that's the report on the number of lags used.

The overall trend rate is actually the Constant in the output since the entire equation is first differenced.

[trangan_20142011]

Dear Tom Doan,

Could I ask you about reading RATS result for LS's structural breaks from the above example? Actually I am newly to RATS program. I used lsunit.rsc to run this test, however I donot how to read the result. Specifically, what does it mean for S(1)? what is the calculated t-statistic to compare with the critical value? How to choose the critical value for model C "Break", as in the example?
Could you recommend me some reference documents for improving my knowledge on using RATS?

Thank you for your attention. I hope to receive your reply asap.
Best regards,

[TomDoan]

Dear Tom Doan,

Could I ask you about reading RATS result for LS's structural breaks from the above example? Actually I am newly to RATS program. I used lsunit.rsc to run this test, however I donot how to read the result. Specifically, what does it mean for S(1)? what is the calculated t-statistic to compare with the critical value? How to choose the critical value for model C "Break", as in the example?
Could you recommend me some reference documents for improving my knowledge on using RATS?

Thank you for your attention. I hope to receive your reply asap.
Best regards,

S{1} is the lagged detrended data---the t statistic on it is the unit root test statistic.

The critical values aren't that sensitive to the location. However, in the example, the estimation range is 7 to 38 and the breaks are at 11 and 31.

D(11) -0.0152 -0.8670
DT(11) 0.0272 2.0811
D(31) -0.0146 -0.7537
DT(31) -0.0119 -1.3742

11 is a bit over 10% into the range and 31 is about 75% in so the closest one will be .2,.8.

If you have a specific textbook that you're using or have used for which we have the worked examples, going through and looking at how RATS does their various examples would be the best way to learn. If you have a certain area of interest that's covered by one of our e-courses (http://www.estima.com/courses_completed.shtml) those can be very useful.

[trangan_20142011]

Hi Tom Doan,
Thank you for your reply.
In the results for model C, there are two calculated values for D(11) and DT(11). What t-statistics (for D(11) or DT(11)) should I use to compare with the the critical value? I ask this because Model C for both intercept and trend stationary. In my study, there is little statistically significant breaks, is there any problem for this?
Many thanks.

[trangan_20142011]

Hi,
Moreover, in terms of text books, I have some, but none of them includes examples about using RATS. Please advise me, tks.

[TomDoan]

Hi Tom Doan,
Thank you for your reply.
In the results for model C, there are two calculated values for D(11) and DT(11). What t-statistics (for D(11) or DT(11)) should I use to compare with the the critical value?

Neither. The test statistic is the t-stat on S{1}.

I ask this because Model C for both intercept and trend stationary. In my study, there is little statistically significant breaks, is there any problem for this?

No. Read carefully section 11.6 of the RATS v8 User's Guide, particularly the second paragraph.

[trangan_20142011]

Dear Tom Doan,

I read the citation that Doan (2000) recommends against differencing even if the variable contains a unit root because it throws away in information concerning the co-movements of variables. Could you explain it for me? If it is right, why do we need to test unit root?

Thank you very much.

[TomDoan]

Dear Tom Doan,

I read the citation that Doan (2000) recommends against differencing even if the variable contains a unit root because it throws away in information concerning the co-movements of variables. Could you explain it for me?

That's specifically in the context of a VAR, and that's not all that controversial---if there is cointegration, then a VAR estimated in differences is misspecified.

If it is right, why do we need to test unit root?

Because referees insist upon it. That's not intended to be flip. That's reality.

[trangan_20142011]

Hi Tom Doan,
Thank you for your reply. I conducted some tests for unit root, and I got different results about stationary for each variable. Can you advise me about what result I should choose? Could I base on the power of test to decide? What should I do next? Thanks a lot.

[TomDoan]

Which tests are consistent with a unit root and which aren't?