problems to use the command keep if

[Elio Rico]

I have a Dataset with 150 countries. Now, I would like to keep the data of only about 35 countries. All the countries are identified by their isocode in the data set.
I tried the following command:

keep if isocode = "AUT"

so this one worked. The problem is, that I want to keep 34 other countries.
I tried to use the operator & ( keep if isocode =" "AUT" & "AUS" &......) but it didn't work out.

Does anyone knows a better way to deal with this problem? I would be extremely thankful!

Best
Elio

[Marcos Almeida]

Hello Elio,

You may want to take a look at this FAQ: http://www.stata.com/support/faqs/da...-observations/

Hopefully that helps.

Best,


Marcos

[Roberto Ferrer]

One way is:

clear
set more off

*----- example data -----

input ///
str3 country
ABW
AFG
AGO
GIB
GIN
GLP
NLD
NOR
NPL
end

list

*----- what you want -----

// -inlist()- works better with numeric arguments
encode country, gen(country2)

// define countries to keep
local tokeep "1, 5, 6"

// keep
keep if inlist(country2, `tokeep')

list

You mean the logical operator OR and not AND, but the original syntax is incorrect anyway. It should be more like:

keep if country == "ABW" | country == "GIN" | country == "GLP"

[Elio Rico]

Thank you very much for your help, Marcos Almeida and Roberto Ferrer. It worked out perfectly! I will definitely have a much closer look at the FAQ next time, sorry about that.
Have a nice day !

Best
Elio

[Nick Cox]

There are problems here at several levels. Roberto has focused on the most important. "Or" logic is needed here, as
"and" logic does not apply. Stata evaluates true-or-false statements within observations, not across them, and an observation cannot be simultaneously for two or more countries.

Further,

 
keep if isocode = "AUT"

would not have worked because it is illegal. The code should have included == not =. That in itself is trivial as most Stata users should be able to spot the typo, but an alarmingly large proportion of the threads here move more slowly than they should because posters do not post the exact code they used.

Note that even something like

 
keep if isocode == "AUT" | "AUS"

would not have worked as this is parsed as if

 
(isocode == "AUT") | "AUS"

and the text "AUS" cannot be interpreted as numeric and so cannot be evaluated as true or false.

[Tobias Wendler]

Good morning everybody,

I`m happy to have found this thread and not opening a new one on a question, which I may resolve but at the moment I`m struggling to find the answer.


I`m having a following variable (string) named Indicator.

dataex Indicator in 153/157

input str213 Indicator
"Domestic extraction (Metal ores) (Tonnes)"
"Indicator name: Domestic extraction (Non-metallic minerals) Data Source: UNEP(http://unep.org/)"
"Description: The amount of non-metallic minerals extracted from the natural environment used in the economy."
"INDICATOR NAME (unit)"
"Domestic extraction (Non-metallic minerals) (Tonnes)"

Now I want to only keep observations in which I have the actual variables and values, so I would want to drop e.g. each one including Indicator name or description, since those have just come into the dataset due to the importing of raw data.

How can I tell Stata to drop any observation where the manifestation of the Indicator variable includes something like description or indicator name and so on.

Thank you so much for support and best regards
Tobi

[Nick Cox]

Perhaps

drop if strpos(lower(indicator), "indicator") | strpos(lower(indicator), "description")

[Tobias Wendler]

Hey Nick,

thank you very much, I finally typed:

drop if strpos(lower(Indicator), "indicator") | strpos(lower(Indicator), "description")

(with indicator written small Stata was not able to find it)

It perfectly worked

Thanks and best regards
Tobi