As a part of my master thesis, I'm conducting an observational study of journalists' background and the effect of this on the approach to a certain case study. I have a fixed number of observations/articles, 781 to be precise. I know that some of them are likely written by the same journalists, but I don't know exactly how many journalists I'll find in the sample. I will conduct a logistical regression.

Can anybody help me with estimating if this is enough articles (power sample size)? I have tried the powerlog in Stata, but I can't account for the cluster-structure arising from having multiple articles from the same journalists.

Please do let me know if I should improve the questioning (as I'm new on this forum).

Best regards, Simon

  • $\begingroup$ Welcome to CV! Some questions: (1) Why do you want to conduct a power estimation if you have a fixed sample size? (2) If you are using logistic regression, what is the outcome? If it is not which journalist, then the repeated journalists will cause some dependence in your data, which you could account for in a mixed model, but not ordinary logistic regression. If it is which journalist, then I could you explain how you intend to classify more than two labels using a binary classifier? $\endgroup$ – Frans Rodenburg Aug 16 '18 at 12:12
  • $\begingroup$ Sorry for not making myself clear: 1) The reason for the power estimation is, that I want to evaluate the feasiblity of the project before conducting it. 2) The outcome is the presence/absence of a specific framing in each article. The independent variables are features of the journalist and the news paper etc. Do that clarify it? Otherwise, please let me know. $\endgroup$ – S. Engfred Aug 16 '18 at 12:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.