Hello!
I am confused between two model choices, hence would like an opinion on what might be appropriate. Here is how my data is setup:
I have 120 firms in my data with daily frequency over 12 years. This translates to a pretty long time-series of a short cross-section. My regression specification is something akin to:
y_t = y_t-1 + x_t + z_t + error
So, basically, the y variable is dynamic by way of the lagged dependent variable. Typically, I would have used the GMM routine of xtabond2 or xtdpdsys. The issue I have is that my data is over 300k large, and trying to do these routines results in instrument proliferation, even after using the maxldep option. I have also done the xtserial test for autocorrelation to confirm its use.
My question build down to this: is the dynamic panel routine of xtabond2 (or xtdpdsys) the only way to go, or should I be using some other technique? Perhaps the xtivreg2 with fd option?
Apologies if my question is a little hard to follow.
Thanks!
Rohit
I am confused between two model choices, hence would like an opinion on what might be appropriate. Here is how my data is setup:
I have 120 firms in my data with daily frequency over 12 years. This translates to a pretty long time-series of a short cross-section. My regression specification is something akin to:
y_t = y_t-1 + x_t + z_t + error
So, basically, the y variable is dynamic by way of the lagged dependent variable. Typically, I would have used the GMM routine of xtabond2 or xtdpdsys. The issue I have is that my data is over 300k large, and trying to do these routines results in instrument proliferation, even after using the maxldep option. I have also done the xtserial test for autocorrelation to confirm its use.
My question build down to this: is the dynamic panel routine of xtabond2 (or xtdpdsys) the only way to go, or should I be using some other technique? Perhaps the xtivreg2 with fd option?
Apologies if my question is a little hard to follow.
Thanks!
Rohit