Looping

[Will Vanlaer]

Hello,

I got a question with regards to the use of loops.
I have panel data and would like to plot the savings rate for each country (using a seperate graph for each country and then combining them in one graph).
Currently, I have the following code.

line savings_rate year if cid=="AT", xtitle("") ytitle("") title("Austria")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\AT.gph"

line savings_rate year if cid=="BE", xtitle("") ytitle("") title("Belgium")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\BE.gph"

line savings_rate year if cid=="BG", xtitle("") ytitle("") title("Bulgaria")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\BG.gph"

line savings_rate year if cid=="CY", xtitle("") ytitle("") title("Cyprus")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\CY.gph"

line savings_rate year if cid=="CZ", xtitle("") ytitle("") title("Czech Republic")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\CZ.gph"

line savings_rate year if cid=="DE", xtitle("") ytitle("") title("Germany")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\DE.gph"

line savings_rate year if cid=="DK", xtitle("") ytitle("") title("Denmark")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\DK.gph"

line savings_rate year if cid=="ES", xtitle("") ytitle("") title("Spain")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\ES.gph"

line savings_rate year if cid=="FI", xtitle("") ytitle("") title("Finland")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\FI.gph"

line savings_rate year if cid=="FR", xtitle("") ytitle("") title("France")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\FR.gph"

line savings_rate year if cid=="HU", xtitle("") ytitle("") title("Hungary")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\HU.gph"

line savings_rate year if cid=="IE", xtitle("") ytitle("") title("Ireland")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\IE.gph"

line savings_rate year if cid=="IT", xtitle("") ytitle("") title("Italy")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\IT.gph"

line savings_rate year if cid=="NL", xtitle("") ytitle("") title("Netherlands")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\NL.gph"

line savings_rate year if cid=="PT", xtitle("") ytitle("") title("Portugal")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\PT.gph"

line savings_rate year if cid=="SE", xtitle("") ytitle("") title("Sweden")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\SE.gph"

line savings_rate year if cid=="SI", xtitle("") ytitle("") title("Slovenia")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\SI.gph"

line savings_rate year if cid=="SK", xtitle("") ytitle("") title("Slovakia")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\SK.gph"

line savings_rate year if cid=="UK", xtitle("") ytitle("") title("United Kingdom")
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\UK.gph"


gr combine "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\AT.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\BE.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\BG.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\CY.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\CZ.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\DE.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\DK.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\ES.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\FI.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\FR.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\HU.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\IE.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\IT.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\NL.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\PT.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\SE.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\SI.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\SK.gph" "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\UK.gph"
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\All_countries.gph" 

I presume there is a more efficient way of doing this, using loops but can't figure out how to do this exactly.
Can someone give a hand?

Thanks,

W.

[Raffaele Grotti]

Hi Will,
this code should serve your purpose.

local countries "AT BE BG"
local cname "Austria Belgium Bulgaria"
foreach cy of local countries {
    gettoken cn cname : cname
    line savings_rate year if cid=="`cy'", xtitle("") ytitle("") ///
        title("`cn'") name("`cy'")
}
gr combine `countries'
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\All_countries.gph"

However, you might to consider to use -by()-

line savings_rate year, xtitle("") ytitle("") by(cid)

Raffaele

[Nick Cox]

Don't loop at all -- is my advice.

Consider this strategy

gen country = ""
replace country = "Austria" if cid == "AT"
* and so on

line savings rate year, by(country) xtitle("") ytitle("")

Clearly there is a bit of work there fixing the country names into a new variable, but the pay-off with by() is worthwhile.

That said, 19 graphs combined are a stretch. Consider

1. whether alphabetic order best satisfies your purpose.

2. whether if comparison between savings and rate is the issue, showing the difference savings - rate as a new variablewould work better.

3. if yes to 2 then group countries as groups of 5, 5, 5, 4 to get a four panel graph that is likely to make comparisons easier.

EDIT: Raffaele made several similar points.

[Will Vanlaer]

Hi Will,
this code should serve your purpose.

local countries "AT BE BG"
local cname "Austria Belgium Bulgaria"
foreach cy of local countries {
gettoken cn cname : cname
line savings_rate year if cid=="`cy'", xtitle("") ytitle("") ///
title("`cn'") name("`cy'")
}
gr combine `countries'
graph save Graph "C:\Users\wvanlaer\Dropbox\Doctoraat\Artikel sparen\Stata\All_countries.gph"

However, you might to consider to use -by()-

line savings_rate year, xtitle("") ytitle("") by(cid)

Raffaele

Hey Raffaele,

Thanks for your reply.
I have tried your suggestion and I have 3 remarks/questions:

• I get an error message saying: "option / not allowed". Could that be due to the fact that you have "///" after

  yttile("")

  ? However, if I simply remove the "///" it doesn't work either. What is wrong with the syntax?
• Could I subsitute

  local countries "AT BE BG .."

  with

  local countries levelsof cid

  ?
• What is the exact use of "gettoken"? I looked it up and understand its general use, but could you briefly explain how it applies here?

Thanks and enjoy the weekend,

W.

[Raffaele Grotti]

Hi,
first of all let me say that the use of - by()- is still the best solution as also Nick pointed out

Concerning your remarks:
1. I don't think it is due to "///". In #4 you typed -yttitle- instead of -ytitle-, so I suggest you to check your code.

2. Yes, you can use -levelsof- but you have to use it in the correct way:

levelsof cid, local(countries)

3. Basically, gettoken splits your string (or list of element) in tokens.
In this case we have a string called "cname" as follows: "Austria Belgium Bulgaria ..."
In the first iteration of the loop, gettoken cn cname : cname takes the first token out from "cname" and stores it in "cn"
Thus, "cn" contains now "Austria" while "cname" contains now only "Belgium Bulgaria"
In the second iteration of the loop "cn" will contain "Belgium" while "cname" only "Bulgaria"
I hope I've been clear enough.

Note, however, that you will have problems in cases where the name of the country is a composite name, such as "United Kingdom"
This is another good reason to prefer the strategy suggested by Nick!

A work around for this specific case could be the following:

local countries "AT BE BG UK IT" // or levelsof cid, local(countries)
local cname "Austria Belgium Bulgaria United Kingdom Italy"
foreach cy of local countries {
    gettoken cn cname : cname
    if "`cy'"=="UK" {
        gettoken cn cname : cname
        local cn "United Kingdom"
    }
    line savings_rate year if cid=="`cy'", xtitle("") ytitle("") ///
        title("`cn'") name("`cy'")
}

Best,
Raffaele

[Will Vanlaer]

Don't loop at all -- is my advice.

Consider this strategy

gen country = ""
replace country = "Austria" if cid == "AT"
* and so on

line savings rate year, by(country) xtitle("") ytitle("")

Clearly there is a bit of work there fixing the country names into a new variable, but the pay-off with by() is worthwhile.

That said, 19 graphs combined are a stretch. Consider

1. whether alphabetic order best satisfies your purpose.

2. whether if comparison between savings and rate is the issue, showing the difference savings - rate as a new variablewould work better.

3. if yes to 2 then group countries as groups of 5, 5, 5, 4 to get a four panel graph that is likely to make comparisons easier.

EDIT: Raffaele made several similar points.

OK, I have also tried this method and it indeed is a bit more straightforward.
Just one question with regards to the axes: now, only the graphs on the left show values for the y-axis, and only the bottom graphs show values for the x-axis. How can I counter this?

[Will Vanlaer]

Hi,
first of all let me say that the use of - by()- is still the best solution as also Nick pointed out

Concerning your remarks:
1. I don't think it is due to "///". In #4 you typed -yttitle- instead of -ytitle-, so I suggest you to check your code.

2. Yes, you can use -levelsof- but you have to use it in the correct way:

levelsof cid, local(countries)

3. Basically, gettoken splits your string (or list of element) in tokens.
In this case we have a string called "cname" as follows: "Austria Belgium Bulgaria ..."
In the first iteration of the loop, gettoken cn cname : cname takes the first token out from "cname" and stores it in "cn"
Thus, "cn" contains now "Austria" while "cname" contains now only "Belgium Bulgaria"
In the second iteration of the loop "cn" will contain "Belgium" while "cname" only "Bulgaria"
I hope I've been clear enough.

Note, however, that you will have problems in cases where the name of the country is a composite name, such as "United Kingdom"
This is another good reason to prefer the strategy suggested by Nick!

A work around for this specific case could be the following:

local countries "AT BE BG UK IT" // or levelsof cid, local(countries)
local cname "Austria Belgium Bulgaria United Kingdom Italy"
foreach cy of local countries {
gettoken cn cname : cname
if "`cy'"=="UK" {
gettoken cn cname : cname
local cn "United Kingdom"
}
line savings_rate year if cid=="`cy'", xtitle("") ytitle("") ///
title("`cn'") name("`cy'")
}

Best,
Raffaele

Thanks for your quick and clear response.

[Nick Cox]

#5 Consider

. sysuse auto
(1978 Automobile Data)

. scatter mpg weight, by(rep78)

. h by option

. scatter mpg weight, by(rep78, xrescale)

The xrescale suboption insists on separate scales. I don't know a way to insist on explicit scales, but identical. If your x range is identical, you may be OK.

[Will Vanlaer]

#5 Consider

. sysuse auto
(1978 Automobile Data)

. scatter mpg weight, by(rep78)

. h by option

. scatter mpg weight, by(rep78, xrescale)

The xrescale suboption insists on separate scales. I don't know a way to insist on explicit scales, but identical. If your x range is identical, you may be OK.

This worked. Thanks for the help!