Each of these implementations of the Pedroni testing procedure produce vastly different results. The user-made command xtpedroni seems much closer to the original papers, producing seven test statistics, 6 of which diverge to - infinity under the null hypothesis. The command xtcointtest pedroni is an official command, but it only produces 3 test statistics, each of which are positive (?), and don't seem to conform to the original papers. Yet it's an official command so I'm inclined to believe it.
Does anybody have any idea on how these two commands can be reconciled?