2
$\begingroup$

First time plotting and interpreting time series data and I have used a line plot for ease of use. I am aware this is incredibly basic, but any input/ recommendations would be much appreciated (e.g., is anything unclear?).

  • My main concern is whether I have adequately displayed the data and
    whether I can do anything useful to improve (e.g., moving average)?
  • Additionally, whether I have interpreted this time series data appropriately:

"The relative frequency of affect-related tokens (counts per 10,000 tokens in psychology abstracts) increased from 3.51% in 1980 to 4.87% in 2017—-an overall relative increase of 39%. The relative frequency of affect-related tokens shows an increase at a rate of approximately 0.037 units per year (over 37 years). Overall, this displays a rapid growth trend in academics' use of emotion-related terms in psychology abstracts over time."

relative frequency of emotion-related words in psychology over time

$\endgroup$
0
$\begingroup$

I am aware this is incredibly basic, but any input/ recommendations would be much appreciated (e.g., is anything unclear?).

In general it is strongly recommended to communicate knowledge in the most simple way possible, as long as it's accurate. Yes a line plot is simple but there's nothing wrong with that, in this case I can't think of any better way to convey the observation accurately.

My main concern is whether I have adequately displayed the data and whether I can do anything useful to improve (e.g., moving average)?

It perfectly shows the increasing trend which is your main point:

  • A moving average would not really help in this case: it's useful only when there's too much noise/variation in the data and the general trend is hard to see. Here you apparently have enough data and/or the data is stable enough, so there's simply no need.
  • You took care of making the plotted value a proportion per 10,000 tokens, and this is also a good idea since an increase in absolute value could be biased: if the total number of abstracts increases, the proportion of abstracts containing the target terms would naturally increase as well even though this doesn't show anything.

Overall this looks very good to me. Just a small note, in your description you could mention that the rate by year is an average: "approximately 0.037 units per year in average".

$\endgroup$
2
  • $\begingroup$ Thank you very much for your feedback! If I may, I would like to ask a few simple questions: 1) I believe I have plotted the value on the y-axis as a proportion per 10,000 tokens. This is the dummy code/ formula: Is it correct? (summed frequency) * 10,000 / total tokens in abstract per year 2) Additionally, is it correct to calculate the average rate by year as follows: slope = rise/run = (4.87-3.51)/(2017-1980) = 1.36/37 = 0.037 $\endgroup$ – n.baes Dec 16 '20 at 2:18
  • $\begingroup$ @n.baes: 1) yes correct 2) correct as well, but I think I would call this value "average increase by year" just to avoid any confusion since the word "rate" could also mean "growth rate", in which case it would be: (4.87-3.51)/3.51=38.7% for the growth rate over 37 years, which is 38.7%/37= around 1% growth rate by year. To be clear your usage of the word "rate" is correct, I'm just mentioning another possibility. $\endgroup$ – Erwan Dec 16 '20 at 11:21

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.