How to fix error "outcome does not vary" for OR and logistic regression

[Anna Gwen]

Dear all, I have been struggling to solve this.

I imported data from excel file. And have a variable "screening" that is entered as "yes" and "no".

I have tried all these commands:
egen screeningnew = group(screening),label

destring screening, generate (screeningnew)

encode screening, generate (screeningnew)

However when i want to calculate odds ratio with another variable "sex"
command:

tabodds screeningnew sex
response variable screeningnew not coded 0/1

How do i fix this? thank you so much

[Joseph Coveney]

You could try something like the following.

assert inlist(screening, "no", "yes", "")
generate byte screeningnew = screening == "yes" if !missing(screening)

You might need to do something analogous with the variable for sex.

[Anna Gwen]

It says
226 contradictions in 226 observations

so it cant work

I think it is the way my variable is coded, is there a way to code the variables?
I have went to data variables manager, to edit label, and this is the default one.
Maybe this is not correct?

[Dirk Enzmann]

Anna, please read the Stata Forum's FAQ (especially Section 12) before posting: Show us the commands you are using (by toggling to the advanced editor and using code delimiters by using the button with the # symbol). Additionally, show us (parts of) your data by using the -dataex- command available at SSC.

The outcome variable should be coded as 0 (for no/false) and 1 (for yes/true), not as 1 and 2. Use -recode- to generate a new variable Whooleypos01 to achieve this:

recode Whooleypos (1=0 "no") (2=1 "yes"), gen(Whooleypos01)

(or recode the values of your variable and modify the label definitions accordingly).