Stacking variables selectively with graph bar

[Marie Christin]

Dear Statalists,

apologies in advance if this question has an easy answer/or was asked before:
I intend to create a bar graph with 4 variables of which I want 2 of them stacked on to each other. So far the codes I have used are:


graph bar (sum) new_deep shallow_to_deep new_shallow shallow_reverse, over(y)

(see also attachment Graph_1)
and then I have been trying to create a "group" for the stacking:

gen exp=0
replace exp=1 if new_deep==1
replace exp=1 if shallow_to_deep==1

to graph:
graph bar (sum) new_deep shallow_to_deep new_shallow shallow_reverse, over(y) stack over (exp)

(see also attachment Graph_2)

However this result is also not quite want I want, as in the second graph the stacking is correct, but I want both bars side by side and one unified x-axis from 1997-2014 (just as in the first graph).

Could anyone point me into the right direction of how to achieve this??
Thanks a lot!!

[eric_a_booth]

It looks like you'll just need to change the order of your over() options to

graph bar (sum) new_deep shallow_to_deep new_shallow shallow_reverse,  stack over(exp) over(y)

Consider the difference in the example code:

sysuse auto, clear
graph bar (sum) turn trunk, over(rep78) stack over(for)
graph bar (sum) turn trunk, over(for) stack over(rep78)

[Marie Christin]

Thanks a lot!

[Zelalem Jaleta]

Dear all support providers, Might be this is simple question for you, but I need support.

I have used Multinomial logit model to conduct analysis for multiple variable. At the same time I was interested to conduct analysis of their marginal effect. Accordingly I did and got good result for first and last variable category. However, for one variable category found in the middles, the result was zero for some independent variable and >1 for some of variable. What does marginal effect zero means? how can I interpreter it and is it possible with zero marginal effect. What about the marginal effect of greater than one. Is it also okey/what?

Thanks in advance.

[eric_a_booth]

Dear all support providers, Might be this is simple question for you, but I need support.

I have used Multinomial logit model to conduct analysis for multiple variable. At the same time I was interested to conduct analysis of their marginal effect. Accordingly I did and got good result for first and last variable category. However, for one variable category found in the middles, the result was zero for some independent variable and >1 for some of variable. What does marginal effect zero means? how can I interpreter it and is it possible with zero marginal effect. What about the marginal effect of greater than one. Is it also okey/what?

Thanks in advance.

You'll want to start a new thread for your unrelated question if you want to get a response.