Generating dummy variables using non-numerical data

[claire macfarlane]

Hello. I am working with survey data where participants are asked if they agree with a statement, in this case whether they consider themselves a public servant. They can choose between five answers. I would like to create a dummy variable PublicServant where 1 = yes and 0 = all other answers. I am not used to working with non-numerical answers and am unsure of the syntax needed. Do I use quotation marks around the potential answers? Is there a way to consolidate this code and say 1=yes and 0=all other answers?

This is my drafted code but I get a mismatch error:

gen publicservant=.
replace publicservant=0 if PublicServant=="[2] No"
replace publicservant=0 if PublicServant=="[3] Don't Know"
replace publicservant=0 if PublicServant=="[-1] No answer"
replace publicservant=1 if PublicServant=="[1] Yes"
replace publicservant=0 if PublicServant=="[-2] Does not Apply"

[Mike Lacy]

Giving you a helpful answer here here is hard without an example of your data. Please re-read the StataList FAQ for new users on using -dataex- to show an example of your data.

It is *possible* here, for example, that your error message has occurred because your variable PublicServant is not indeed a string, but that content such as "[2] No" is actually a value label assigned to it. Also, your long series of statements to create a dummy variable is almost certain unnecessary, and can probably be done in one line.

[Al Perez]

gen publicservant = PublicServant == "[1]Yes"

or

gen publicservant = PublicServant == 1

the first one works if Public Servant is a string. the second works if Public Servant is numeric with labels assigned to certain values. you can use

des PublicServant

to find out if the variable is string or numeric.

[Joro Kolev]

As Mike said, without you providing a data sample using -dataex-, it could be anything.