Hi,
I'am currently struggling to test for weak instruments when conducting a 2sls regression using robust standard errors (vce robust). My model ist just-identified (i.e., one instrument and one endogenous variable).
estat firststage provides me with the robust F statistic. However, I do not know to what compare it to. Does the threshold of 10 apply here?
Using weakivtest provides me with an effective F statistic that equals the robust F statisic since my model ist just-identified.
Using ivreg2 provides me with weak identification tests:
My main question is:
I'am currently struggling to test for weak instruments when conducting a 2sls regression using robust standard errors (vce robust). My model ist just-identified (i.e., one instrument and one endogenous variable).
estat firststage provides me with the robust F statistic. However, I do not know to what compare it to. Does the threshold of 10 apply here?
Using weakivtest provides me with an effective F statistic that equals the robust F statisic since my model ist just-identified.
Using ivreg2 provides me with weak identification tests:
- Cragg-Donald Wald F statistic: is this statistic valid when using vce robust? What can I compare it to? Does the threshold of 10 apply here?
- Kleibergen-Paap rk Wald F statistic: this statistic again equals the robust F statistic.
My main question is:
- Does the rule of thumb robust F statistic / effective F statistic / Kleibergen-Paap rk Wald F statistic > 10 apply in this context?
