Testing for autocorrelation in a logistic regression analysis with panel data

[Marco Liedecke]

I am trying scrutinize the relationship between regimetypes (political regimes eg. closed autocracy, hybridregime and liberal democracy) and the risk of civil war onset.

My data spans from 1946-2006 and includes all independent countries. It is thus panel data and therefore I wish to test for autocorrelation.

My understanding is that in order to test for autocorrelation in Stata you use the command xtserial after the regression. I used the following procedure:


. xtset country_name_numeric year
panel variable: country_name_numeric (unbalanced)
time variable: year, 1946 to 2006, but with gaps
delta: 1 unit

. xtlogit onset i.v2x_regime, or

. xtserial onset i.v2x_regime
command xtserial is unrecognized
r(199);

What am I doing wrong since Stata does not recognize the command xtserial?

//Marco Liedecke

(FYI I am new to the forum)

[Carlo Lazzaro]

Marco:
welcome to this forum.
As per FAQ, please note that you have to report the use of a community-contributed Stata programme, such as -xtserial-.
You did nothing wrong, but overlooking that the -helpfile- that comes with -xtserial- states that it works for linear panel data only.

[Marco Liedecke]

Thank you, Carlo

Seeing that my dependent variable is binary (no civil war = 0 and civil war = 1) and consequently I wish to run a logistic regression can you point to any commando that will allow me to test for autocorrelation?

//Marco Liedecke

[Carlo Lazzaro]

Marco:
Tricky issue indeed (https://www.statalist.org/forums/for...utocorrelation).
A basic fix can be to run your regression model with both default and clustered standard errors and see if any difference creeps up between them.

[Marco Liedecke]

Thank you for your help, Carlo

If you have any other sources concerning autocorrelation in regards to logistic regression please link

//Marco Liedecke

[Carlo Lazzaro]

Marco:
I simply googled with the following key-words: -autocorrelation xtlogit-.