Obtain p value

[Katie Holzer]

Hello,
Is there an option I can specify that would provide me the p-value for the relationship between these two variables using the following command (assuming the following command is appropriate in the first place)? This is survey data, overall use is a binary variable 0=same, 1=increasing. Burnout_score is a continuous variable.

bysort overall_use: ci mean burnout_score

[Mike Lacy]

Neither -bysort- nor -ci- is going to be of any use for you here, as neither one calculates any measure of relationship. (That can be discovered from -help ci- and -help bysort-.) So, the question is, what measure of relationship would you like to use? When you have decided that, pointing you in the right direction would be easier. One conventional approach would be a two-independent groups t-test; another would be to (more or less equivalently) use a regression model with overall_use as an indicator variable predictor. See help -ttest- and -help regress-. Now, as you may be aware, hypothesis testing is generally derogated these days in favor of confidence intervals, so using the CIs from -regress- or -ttest- would for many of us be preferable to a p-value.

[Katie Holzer]

Thank you, Mike. I was incorrectly trying to find the unadjusted and adjusted means using bysort.

After looking at the help documents you referenced, I believe the following coding will give me the unadjusted and adjusted means for the variables of interest. If you're able I appreciate your feedback.

///Unadjusted
regress burnout_score i.overall_use
margins overall_use, cformat(%6.2f)

///Adjusted
regress burnout_score i.overall_use i.clinical_role
margins overall_use clinical_role, cformat(%6.2f)

[David Radwin]

If by "survey data" you mean complex survey data (rather than a simple random sample), please see this FAQ: https://stats.idre.ucla.edu/stata/fa...th-survey-data