I have a question regarding the validity of the methods behind this graph I created. enter image description here

On the Y-axis we have objectively measured food intake (daily energy intake) and on the X-axis we have "hunger scores" measured on a scale with a range of 0 to 20 (score of "1.5" not possible...)

As expected, hunger scores positively associate with food intake such that individuals who were hungrier ate more. However, I am not sure if this is to correct way to present this Figure given that many (20%) of individuals have a score of "0" (but the scores are on a ratio scale). Even though Pearon's R is significant I opted for Spearman rho for reasons unclear to me (although I know Spearman is most useful for rank-ordered data in which case this is).


I think this is a fine way to represent your data. Having 20% of the data at hunger=0 is not a problem. Hunger is an independent variable and it does not need to be normally distributed. Showing the trend line clearly shows the positive correlation between the two variables. In addition to reporting a correlation metric (Spearman is fine if just worried about a monotonic relationship rather than a strictly linear one, but I would default to Pearson here since showing a linear plot), I believe some value could be added by reporting the regression coefficient (slope) and intercept of your trend line. Along with these you may report the standard error. Most statistical software packages report these by default. Also, it is good to plot the residuals and show whether they are heteroskedastic as well as plotting a histogram of the residuals to show they are normally distributed. This is important because linear regression makes these assumptions when fitting the data. All of these extras depend on your audience, and what you want to convey to them. For an academic crowd it might be good to go the extra mile.

  • $\begingroup$ Excellent point about monotonicity will opt for Pearson and also include regression coefficient. $\endgroup$ – mindhabits Jul 31 at 7:27

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.