You can chose any software to perform lag selection. i have found some studies for your reference. hope this help

https://www.unine.ch/files/live/sites/irene/files/shared/documents/Publications/Working%20papers/2017/WP17-03.pdf

http://www.eviews.com/help/helpintro.html#page/content/panelstats-Panel_Causality_Testing.html

https://www.youtube.com/watch?v=kbUehCU9BLw

1. VAR is rather different approach than panel. If some lag is significant in VAR framework it does not mean it shold be inclu)ded in panel approach

2. AIC penalizes for more complex models, i.e., models with additional parameters. Like AIC, BIC uses the optimal loglikelihood function value and penalizes for more complex models, i.e., models with additional parameters. The penalty of BIC is a function of the sample size, and so is typically more severe than that of AIC ( Box, G. E. P., G. M. Jenkins, and G. C. Reinsel. Time Series Analysis: Forecasting and Control. 3rd ed. Englewood Cliffs, NJ: Prentice Hall, 1994.)

3. About comand in stata: the choice of the number of lags is rather arbitary, so in a sense you could use some info from results to go back and set different number of lags

4. With yearly data - you shold probably stick with 1 or 2 lags - there is a lot of theretical mambo jambo about the memory of the process iteself, but see how it goes with 1 and 2 lags - it is hard to assume that something that happend 3-4 years ago is so significant.

Thank You Muhammad, and Bojan

So, If the variables are stationary at level, I can use either VAR or HD-Granger Causality.

As you Dr. Bojan mentioned before, I already use different lags to test the causal relation between X, and Y. however, the outcomes have no consensus on the results:

for instance: when using Panel Granger causality ( Dumitrescu and Hurlin 2012 ). the results was:

X↔Y (lag 1)

X↔Y (lag 2)

X↔Y (lag 3)

While when use PVAR, the results was :

Y→X (lag 1)

X→Y (lag 2)

X↔Y (lag 3)

Note: lag length criteria based on var test was 3 lags based on BIC

Therefore, I was confuse which lags and method I should use (HD-Granger causality, or PVAR). However, previous research on such relation also has no consensus on the way of causality.

thank you, and sorry for long text.

Ng and Perron (2001) showed the advantage of MAIC compare to AIC, BIC and MBIC.

Ng, S., Perron, P., 2001. Lag length selection and the construction of unit root tests with good size and power. Econometrica 69, 1519-1554.