[Mike Gonzalez]

Dear Joao:

At Linh Nguyen example what would be the interpretation for grade (asumming it is a dummy variable indicating whether o nor have a college degree)?

Thanks,

[Joao Santos Silva]

Dear Mike Gonzalez

In that case, it means that on average having a degree increases the probability of being unionised by 6.4%. Does it make sense?

Best wishes,

Joao

[Mike Gonzalez]

Yes, it does. thanks a lot Joao. I really appreciate it.

Best,
Mike

[Samuel Lux]

Dear Joao,

I can't use the command as it says too many values.
Is there a limit for values? (I have about 130,000) And is there any solution for this issue

Thanks

Best
Samuel

[Joao Santos Silva]

Dear Samuel Lux,

Thanks for this. I believe I know what the problem is. Please contact me by email and I'll try to provide an updated version.

Best wishes,

Joao

[Samuel Lux]

Dear Joao,

Thank you so much. I sent you an email.

Best
Samuel

[Joao Santos Silva]

Thanks to Kit Baum, a new version of the command is available in SSC; the new version solves the problem that Samuel Lux identified. Thanks, Samuel!

Best wishes,

Joao

[Samuel Lux]

Dear Joao Santos Silva

I have one more question. How would you interprete values which are above 1 ? Do they even make sense?

I am a little confuse about that. For instance, in Linh Nguyen table there is value above one (i.south i think).

Thnak you

Best

Samuel

[Malte Mueller]

Dear Joao Santos Silva,

if I understand correctly, the coefficient for grade in #3 by Linh Nguyen is interpreted the following: If grade increases by 1 unit, the average probability of being unionized goes up by 6,4 percent.

I am wondering, whether it is possible to translate the percentage interpretation into a percentage point interpretation? Since the average of union is 22,2 percent in that sample, can we just say that we have a 1,4-percentage point increase (22,2 percent x 6,4 percent) from 22,2 percent to 23,6 percent, if grade increases by 1 unit?

I am asking, since I want to compare my aextlogit results with another estimation that displays average marginal effects, that show percentage points.

Best,
Malte

[Joao Santos Silva]

Dear Malte Mueller,

That is a ratio of two averages and it is difficult to interpret. I guess you can use it as indicative, but not more than that.

Best wishes,

Joao

[Michelle Eilers]

Dear Joao Santos Silva,

Thank you for the clear documentation on the aextlogit command. I am using this command to fit a model predicting weekly sexual intercourse, with a 2-way interaction. Is it possible to plot the interaction output as you would with the margins command?

My code is the following:

 global margin1 "c.rpvgasxbfm##c.vdssxny_comp marstat_nophys tv_enroll2 tvlivbrthnum tv_fapprovesex tv_papprovesex"

aextlogit rjsex $margin1 if marstat_ud!=1 & pregstat_ud!=1 & numjrnl>=3, nolog

Output for the model:

Many thanks!
Michelle

[Joao Santos Silva]

Dear Michelle Eilers,

It looks as if the command you are using does not have the intended behaviour. Can you please contact me by email so that we can clear that? I can then post here the outcome of our discussion.

Best wishes and thanks,

Joao

[Joao Santos Silva]

Dear Michelle Eilers,

Thank you for contacting me.

I am afraid you cannot do anything like the plot you want when you estimate a model with fixed effects, because in that case it is not possible to estimate the probabilities of the events. Even with random effects, the value of those plots is limited because I believe they show probabilities computed with the random effect evaluated at zero and that is not very informative (just think of Jensen's inequality).

Your post also highlighted that the current version of the command has a small bug; it does not affect the results but produces a very ugly output. I'll ask Kit Baum to update it.

Best wishes,

Joao

[Michelle Eilers]

Thank you for your assistance!

[Joao Santos Silva]

Dear All,

Thanks to Kit Baum, this aextlogit has been updated to eliminate the but noted in #28; many thanks to Kit for this and thanks to Michelle Eilers for bringing it to my attention.

Best wishes,

Joao