r(mean) , panel subgroup

[Rochelle Zhang]

Dear Statalisters,

I saw this old forum discussion , my question is about understanding the code :

original forum list is here
https://www.stata.com/statalist/arch.../msg01036.html


Johanna asked the question

I have a panel of European regions over time. I would like to plot a
variable against time for these regions grouped by countries - that is, I would
like to have all the regions of one country in one plot, and the plots of all
countries together on the same overall graph.

Nick helped him/her out with a code sketch here:

egen country2 = group(country) 
sum country2, meanonly 

forval i = 1/`r(mean)' { 
    xtline response time if country2 == `i' , overlay name(gr`i') 
    local graphs `graphs' gr`i'
} 

graph combine `graphs'

i know Nick uses egen to generate country code . I think the sum function meanonly, will generate the mean of the country2 variable.

e.g. original data has 3 countries, then

sum country2, meanonly

gives 2 ?

my question is about what value goes in to the forval loop, is it the average 2 from above? without the data of the orignal poster and my limited stata knowledge, I am not sure about my interpretation.

Thank you in advance !

forval i = 1/`r(mean)'

Rochelle

[Nick Cox]

Sorry; the result used should be r(max) not r(mean).

[Rochelle Zhang]

Hi Nick, thank you for the correction !

why do you use

 
 sum country2, meanonly

would it be the same effect if I use

sum country2, d

if the goal is to feed the r(max) to the loop ?

[Nick Cox]

Look at the manual entry for summarize. All I want is the maximum. None of the extra stuff detail calculates is of interest. There is no point even to a plain summarize.

[Rochelle Zhang]

thank you, Nick.

From manual, it says:

meanonly suppress the display; calculate only the mean;


what I am confused about is the loop requests

r(max)

so you are saying

 
  sum country2, meanonly

still calculates r(max) and keep it in memory, right?

[Carlo Lazzaro]

Rochelle:
strange as it may seem (LOL), Nick is right, as you can see from the following toy-example:

. sysuse auto.dta
(1978 Automobile Data)

. sum price, meanonly

. return list

scalars:
                  r(N) =  74
              r(sum_w) =  74
                r(sum) =  456229
               r(mean) =  6165.256756756757
                r(min) =  3291
                r(max) =  15906

. di r(max)
15906

.

[Nick Cox]

You're right that the manual doesn't explain this very well. https://journals.sagepub.com/doi/pdf...867X0700700311 was an attempt to do better.

[Rochelle Zhang]

Thanks Nick again ! Also, many thanks to Carlos for the code sketch !

I tried the following code (post #1 above) for my panel data - replacing the variable names "response time" with my variable of interest x and year

sum cpair, meanonly 

forval i = 1/`r(max)' { 
    xtline x year if cpair == `i' , overlay name(gr`i') 
    local graphs `graphs' gr`i'
} 

graph combine `graphs'

option overlay may not be specified with multiple variables

This is my first time using xtline, I searched some posts for discussion on overlay, but did not know how to fix this.

My data : company annual observation across time, the panel is multiple firms multiple years, x variable is firm-year specific. cpair takes 0 or 1 value that stratify the sample into 2 subgroups.

Sincerely,

R

[Nick Cox]

The error message explains itself. The syntax is -- from the help --

xtline varname [if] [in], overlay [overlaid_options]

and in this flavour of the command, and generally, you need not (must not!) specify the time variable.

[Rochelle Zhang]

Thanks Nick. I thought your original posting was using panelid (response) and yearid (time) . I misunderstood it.