Dear community

I am trying to estimate a model using cs-ardl method (xtdcce2). The model has to be dynamic and it is highly possible that there is endogeneity problem, because the expected sign is positive for the lagged dependent variable but I get a negative estimated coefficient. So I try to use iv such that

xtdcce2 d.y (l.y=y) if(year>1986), lr((d.l.y) (l.d.z) (l.v) (d.l.q)) lr_options(ardl) cr(_all) fullsample

and it returns with this error

<istmt>: 3200 conformability error

When I remove "(l.y=y)", the estimation results appear but the results suffer from endogeneity.

In coıntrats, when I try estimating the same model using cs-dl model, the iv code works perfectly and the estimation produces the expected results for all the variables.

Does anyone have a suggestion regarding this? I tried using various lengths of lags, differencing variables and reducing the number of explanatory variables but it did not help.