This table shows the number of stories reported by the New York Times on Afghanistan in which a US government official was quoted.

Each row does not represent all stories but, rather, a sample of three constructed weeks of stories during that time period (IOW, all Afghanistan stories from three randomly selected Mondays, three randomly selected Tuesdays, etc.).

enter image description here

Would it be safe to simply say "overall, the New York Times quoted American government officials less than half of the time in its reports on Afghanistan during the three time periods"?

Or, should I (or could I even) also report a margin of error?

  • Were I to report a margin of error I don't know the actual population. Could I safely use a predicted population of 784 (since I'm using a sample of three constructed weeks to report on three months of content - or nine weeks to report on nine months overall - and I have 98 stories in that sample)?

  • Constructed week sampling is a form of stratified random sampling and it's been established elsewhere that three constructed weeks can be used to generalize six months of newspaper content, which is the reason I have three constructed weeks for each time period (more here: https://cphss.wustl.edu/Products/Documents/Papers_Luke_et_al_2011_How_much_is_enough_New_recommendations_for_using_constructed_week_sampling.pdf). Given that, would layering on something like a margin of error even be appropriate or should the numbers be taken "as is"? Can margins of errors be applied to non-human surveys?

  • Is there a different tool I should be using?

Thanks, in advance, for any suggestions or assistance!

  • $\begingroup$ Are you sure the three are comparable? One looks very different. $\endgroup$ – mdewey May 29 '17 at 19:36
  • $\begingroup$ Sorry, I perhaps worded my question poorly. I'm not actually trying to compare the three, I'm just trying to see if it would be reasonable to say "during the first six months of 2014, 47.5% of New York Times stories about Afghanistan contained a quote from a US government official," etc., without qualification. Or, under these circumstances, would it be appropriate to describe a margin of error to qualify this 3-week SRS sample and how would that be applied in the absence of absolute certainty of the population of stories from that time period? $\endgroup$ – Ev Ze May 29 '17 at 23:12
  • $\begingroup$ You are assuming they all estimate the same thing but I doubt that. $\endgroup$ – mdewey May 30 '17 at 15:34
  • $\begingroup$ I'm sorry, I don't understand your comment, mdewey. These aren't estimates, the above chart shows measured results from three constructed weeks. $\endgroup$ – Ev Ze Jun 1 '17 at 0:20

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.