Odd issue when using xtcd and xtcd2 to test for cross sectional dependence

[Nicole Moskowitz]

Dear collegues,

I am in Stata 17 and two weeks ago I installed the xtcd and xtcd2 packages in a laptop and 3 days ago I did the same in a desktop computer. In the laptop, when I type

xtcd gdp
xtcd2 gdp, noestimation

I get the obvious different results (216.25 and 453.436, respectively) as the first line of code calculates the Pesaran's 2004 statistic and the second line the Pesaran's 2015 statistic. However, when I type the same commands in the desktop computer, the results are exactly the same after using both commands (216.25, in other words the result of Pesaran (2004) test). When I use the xtcdf package, I get the same result as the xtcd package, as expected. Both computers have even the same operating system (Ubuntu 20.04). I tried with other variables and the problem remains the same. Although unthinkable for me, my only guess is a possible conflict or something between xtcd and xtcd2 due to a recent update of xtcd2. But I really believe that is almost impossible. Has anyone experienced recently same odd situation? If it is not too much to request, can anyone quickly update both packages and check if they are facing the same problem?

Kind regards,

Nicole

[Nicole Moskowitz]

Just a correction to the post, when I wrote "obvious different results", I meant that I have used Stata to test for cross sectional dependence for few months and I have always obtained different test statistics.

[Andrew Musau]

These are not official Stata commands, so the universe of those who can reply positively to your questions is limited. You are better off sending an email to the authors to address the issue. What usually explains different results is different versions, which you can check by typing

which xtcd
which xtcd2

If the versions are different, then that explains it. If you have the same versions, and you are running the same model in both computers and getting different results, then it may be a bug with the program. But it would not be unreasonable that the authors have updated reporting of the test statistics, for one reason or another.

[Tasfaye Fikadu]

Dear collegues,

I am in Stata 17 and two weeks ago I installed the xtcd and xtcd2 packages in a laptop and 3 days ago I did the same in a desktop computer. In the laptop, when I type

xtcd gdp
xtcd2 gdp, noestimation

I get the obvious different results (216.25 and 453.436, respectively) as the first line of code calculates the Pesaran's 2004 statistic and the second line the Pesaran's 2015 statistic. However, when I type the same commands in the desktop computer, the results are exactly the same after using both commands (216.25, in other words the result of Pesaran (2004) test). When I use the xtcdf package, I get the same result as the xtcd package, as expected. Both computers have even the same operating system (Ubuntu 20.04). I tried with other variables and the problem remains the same. Although unthinkable for me, my only guess is a possible conflict or something between xtcd and xtcd2 due to a recent update of xtcd2. But I really believe that is almost impossible. Has anyone experienced recently same odd situation? If it is not too much to request, can anyone quickly update both packages and check if they are facing the same problem?

Kind regards,

Nicole

Nicole Moskowizt,

I am doing on panel data with N<T, I want to do cross sectional dependence test. However, I cannot do pesaran test. I have to do Breusch- Pagan cross sectional dependence test. I found xttest2 code. But, this is after fixed effect ( May be I am wrong). I am doing on pooled mean group ARDL model. Did you come across stata command for Breusch- Pagan cross sectional dependence test?