running VEC, problem in lrtest

[varun miglani]

i ran vec command

quietly vec ln_price_2 ln_price_1 ln_price_3 ln_price_4 ln_price_5 ln_price_6, trend(constant) rank(5) lags(3)
estimates store constantlags3
quietly vec ln_price_2 ln_price_1 ln_price_3 ln_price_4 ln_price_5 ln_price_6, trend(constant) rank(5) lags(2)
estimates store constantlags2
lrtest constantlags3 constantlags2


*last command not executed, observations differ: 147 vs. 148.why am i getting error r(498)
i have stata 14, how come observations differ when i run the same vec model with two lags and

[Carlo Lazzaro]

Varun:
the difference in the e(sample) observations make sense if you consider the mechanic of the L. operator:

. encode make , g(num_make )

. tsset num_make

Time variable: num_make, 1 to 74
        Delta: 1 unit

. g A=L2.price
(2 missing values generated)

. g B=L3.price
(3 missing values generated)

. sum price A B

    Variable |        Obs        Mean    Std. dev.       Min        Max
-------------+---------------------------------------------------------
       price |         74    6165.257    2949.496       3291      15906
           A |         72    6074.778    2907.061       3291      15906
           B |         71    6094.183    2923.052       3291      15906

.

[varun miglani]

in the book introduction to time series using STATA by Sean Becketti, he uses the command lrtest rconstantlags3 rconstantlags2 after storing the estimates of these vec
the objective is to check, if we can trim the lag. as Each lag adds coefficients. Since i have 6 variables, how do i decide on lags. is 2 lags enough, or should i go for 3 lags, will it depend on P>|z| . request for guidance

[Andrew Musau]

Assuming that the underlying logic of this approach is correct, you can use the -force- option of lrtest to override this error.

Each lag adds coefficients.

This assumes that the model with a lower number of lags is nested in the model with a higher number of lags.

lrtest constantlags3 constantlags2, force

[varun miglani]

thanks, the query is resolved