0
$\begingroup$

I'm plotting time series data where the 5th-95th percentile values overlap so it looks pretty messy. The data is based on a simulation which tracks military 96,000 military personnel purchase of clothing, which is based on a point system. What I'm plotting in the upper chart is the total points used by all personnel in each year from Year 1 (FY 17/18) to Year 7.

To make the charts readable I just used the means in a plot, which looks like this:

enter image description here

You see the mean of S8 rises from year 1 to year 2 and stays nearly constant for the next 5 years.

I thought it would look better if I just plotted the data as a distribution, but I wasn't sure if this made sense given that it's no longer a time series. The chart I got looks like this:

enter image description here

Sample Data for the line and dist charts are as follows:

S8Line <- structure(list(X1.23421911 = c("2 23565938", "3 23583827", "4 23627453", "5 23608291", "6 23600443", "7 23590607", "1 23417838", "2 23619367", "3 23633472", "4 23623865", "5 23599065", "6 23589071", "7 23642366", "1 23433516")), row.names = c(NA, 14L), class = "data.frame")

S8Dist <- structure(list(X1.23421911 = c("2 23565938", "3 23583827", "4 23627453", "5 23608291", "6 23600443", "7 23590607", "1 23417838", "2 23619367", "3 23633472", "4 23623865", "5 23599065", "6 23589071", "7 23642366", "1 23433516")), row.names = c(NA, 14L), class = "data.frame")

$\endgroup$
  • $\begingroup$ The relationships among your plots are obscure: could you please describe your data and explain how the bottom plots reflect the top plot? $\endgroup$ – whuber Mar 25 at 21:48
  • $\begingroup$ I added some info in the OP. $\endgroup$ – Angus Mar 25 at 21:54
1
$\begingroup$

Info conveyed by the two plots are very different.

One shows the behavior of the mean purchase in each category over time. The take away message is that after a couple of years purchases in each category stabilise around a mean value, and these might be different for each category (or not, we can't know without some form of error bands or statistical significance test).

The second plot ignores the time dimension and gives the distribution of purchases in each category.

Which one you need depend on your goal in the analysis. If you are interested in performance over time then is the first one, if you do not care about the time, and frequencies and spread of purchases are relevant then maybe the second.

Answer the question for yourself: what is the point of the plot? What information are you trying to convey?

| cite | improve this answer | |
$\endgroup$
  • $\begingroup$ Thank you. In the past members were not using all their points and so there was an excess of points which when translated into a dollar figure added a high financial risk element. I adjusted limits in the simulation to show the impact on usage and excess points at the end of each FY. My feeling is that the second graph is more telling on the new limits, which are less than the old limits. In any case, I can generate both types and place the time dimension charts in an annex. $\endgroup$ – Angus Mar 26 at 1:01

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.