Closed apascualgarcia closed 3 years ago
Box plots figures do not render well in the panels I can prepare a couple of options and then we can decide. What is the problem with the boxplots?
I need to know if you wil generate also the remaining figures going into the SM. I happy to generate them too.
Please let me know when you think this will be ready. I will work on this for the whole day on Monday. It should be ready by the end of Monday.
Please have a look at the current version of the Ms. Since we are creating panels with several figures the most important thing is that it is possible to clearly see the trends when parameters are changing, and it is very difficult to see any trend in a small box plot filled of dots. A nice possibility is the type of plots of the paper of Zandvoort, where the emphasis is on the trends and the variability are just shadows.
This could actually be an easy possibility, I could extract the IQ instead of mean and stdv with my script and plot the mean and the IQ limits with a shadow
Box plots figures do not render well in the panels, change to something like geom ribbon or similar: it is needed to see the trend. Perhaps the vio with a geom line added could work, it is cleaner than the dots
I generated a bunch of combinations for the designs. Personally I prefer violin plots without extra lines. I am slightly against lines because the x-axis is not an actual scale, but factors.
Isolation Figures are wrong. For number of deaths if there are no tents it should be around a 10%
It is fixed now.
Please include all the values simulated for the number of tents, as it is shown in the figure for the probability of outbreak.
Done.
Exclude simulations with no deaths from the statistics. Also from the boxes, e.g. in the vio figures seem to be included, if anything it would be better to write the number of simulations with no deaths at zero, see e.g. this figure, but this number is 1-Poutbreak so it would be redundant, although perhaps for some people it helps.
Simulations with no deaths are not included. The data points towards the bottom of the boxplots are simulations with a very low number of deaths, but strictly positive. All the plots (old and new, box and violin) are generated from the same summary table.
The time to peak for isolation shows fraction of deaths, the name of the file is actually wrong and the one expected does not exist.
Fixed.
Make all figures uniform, e.g. background theme, legend names and colours
Done. They are all generated from the same main script now, so it is easy to do changes.
Increase all fonts, remove titles.
Fontsizes are fixed at the beginning of the script. I can make them bigger or smaller easily, once we see how it looks with the current configuration in the manuscript.
If it is possible, create directly a panel with all figures to have a single legend and grid/wrap titles.
Done.
Thanks @ecam85 the panels now look great!
I generated a bunch of combinations for the designs. Personally I prefer violin plots without extra lines. I am slightly against lines because the x-axis is not an actual scale, but factors.
I see your point, but I think it helps to see if the means/medians increase or decrease along the parameters, we are not claiming anything about the functional forms, if the line is thin I think it is clear that is just to help the reader. Actually violin plots are for me the most difficult to see the differences. I am inclined now towards boxplots with the lines or one of the ribbons, so perhaps we can gather more opinions?
@Jennifer-Villers, @Chamsy-Sarkis and @jordan-klein, could you please have a look at the figures labelled _col.pdf
in this folder and say which ones you think are more clear? (they are different versions of the same six figures)
Eduard, some minor aesthetic modifications:
We will indeed need larger font sizes, I think in the text the figure will be very similar to what we see online.
Done.
Change exposed --> exposed zone, shielded --> safety zone (we do not longer use "shield" to avoid misunderstandings with isolation centers)
Done.
Change colours: safety zone --> green; exposed zone --> orange; all --> maybe grey? Or blue?
I change it to safety (green), exposed (orange), all (blue).
Perhaps adding the title of the intervention in the top of the column?
Done!
I also increased the size of the "trend line" in the boxplots. I like boxplots too, rather than the ribbon plots in this situation.
I renamed the plots by "rows" to old, I updated the ones by columns (*_col.pdf).
@Jennifer-Villers, @Chamsy-Sarkis and @jordan-klein, could you please have a look at the figures labelled
_col.pdf
in this folder and say which ones you think are more clear? (they are different versions of the same six figures)
Hi Alberto and Eduard, I had a look at the figures as you recommended. Regarding clarity, I would privilege either the boxplots (with a slight preference for the ones with lines as it makes it easier to see what you need to compare) or the ribbon with standard error. The violin plots are extremely distracting (maybe it is because I've never used them before but my eyes are attracted to the various shapes and not at all to their meaning, and I am sure some readers may have the same problem). I hope this helps!
@apascualgarcia @ecam85 I am also inclined towards boxplots with lines between them, so either boxmean_col.pdf or boxmedian_col.pdf
Fix figures (see current appearance in the pdf of the MS)
I need to know if you wil generate also the remaining figures going into the SM.
Please let me know when you think this will be ready.