dannyparsons
commented
6 years ago I think there are two simple changes that could be done now:
-
Instead of a checkbox for exceedance have two radio buttons at the top for cumulative/exceedance. This would then be consistent with other dialogs like boxplot
-
Since the y axis is always 0 to 1 I suggest we have the break point options on the main dialog. So the sequence from, to, by controls. I think this would also make the code easier since it is
ggplot2::scale_y_continuous(breaks=seq(from=0, to=1, by=0.1))for cumulative andggplot2::scale_y_reverse(breaks=seq(1,0,-0.25), labels = seq(0,1,0.25))for exceedance. This is already hard coded into the dialog for exceedance so now having it optional shouldn't be difficult.
These are both good tasks for someone for someone wanting to get into understanding ggplot2 and R code in general to support Maxwell on graphics.
@rdstern will follow up more generally on graphics I think.
maxwellfundi
rdstern
Ogik99
I needed to do these for PICSA in Tanzania. They are important and can look good in R (and hence R-Instat). The dialogue is in Describe > Specific > Cumulative Plots and also in Climatic > Seasonal Forecast Support > Cumulative/Exceedance Graphs. Once it is working a bit better, then in Climatic I suggest we move it (or possibly add it also to the Climatic > PICSA menu..
I have suggested Maxwell and Danny, not because they will fix the problems. But it is a good time to consider who could be looking at the plotting with this possibly as a first example. This will need supervision by Maxwell, and some edits, mentioned below, could be easy to do.
Here are 2 examples:
They are the opposite of what statisticians call cumulative frequency graphs. The Exceedance graphs go down from 1 to zero and the probability (left hand side) shows the chance of more than the given x value.
It is possible that the improvements needed are general, rather than just this dialogue.
a) When I change the y-axis options to Continuous I find it is initially set to discrete. (I assume this setting is somewhere in the code for this dialogue, rather than general? It is an interesting issue, because the y variable is sometimes an integer column (like length in complete days), but the y-scale in the graph is from 0 to 1, so is always continuous. b)The default on the y-variable is to give me labels at 0, 0.25, etc and the request (in PICSA is for more labels to help read off the corresponding values more easily. I change it to numeric and make the necessary edits. It ignores me and when I return it is reset to discrete again. (I had to edit the script file to get the 0.1 spacing in the figures above. The problem is clear from the code below. c) It is because the R code seems contorted and repetitive. I give an example below. It shows the settings of 0.1 were accepted, and then overwritten by later code. d) The Line options in the dialogue are disabled. That stops me being able to change the colours, thickness etc. e) Curiously the main options - please call it Plot Options (rather than just Options) were enabled. I say curiously, because I was using multiple Y variables (to get the 2 lines in the examples above) and on other plots, the sub-dialogues are disabled. I like it enabled! f) We still have the general problem that when we add the points - which we can do, from the main dialogue - and I like that - we can't change the options on the points layer, because they are from the main dialogue. This is a general problem that is reported elsewhere and has been an issue for a long time. g) What I am not sure if it is easy is to change the y scale. There is the option in the dialogue - which is good, and that is to have counts, rather than proportions. Is it easy to also 1) Have percentages, and 2) Be able to multiple by 10 rather than 100 - so it gives the risks in 10 years, rather than in 100!
Here is an example of the current code from the command:
Code generated by the dialog, Cumulative/Exceedance Graphs
merge2 <- data_book$get_data_frame(data_name="merge2", stack_data=TRUE, id.vars=NULL, measure.vars=c("sum_Dodoma","sum_Kondoa")) last_graph <- ggplot2::ggplot(data=merge2, mapping=ggplot2::aes(x=value, colour=variable)) + ggplot2::stat_ecdf() + theme_grey() + ggplot2::geom_point(stat="ecdf") + ggplot2::scale_y_continuous(breaks=seq(from=0, to=1, by=0.1)) + ggplot2::scale_y_continuous(breaks=seq(by=0.1, to=1, from=0)) + ggplot2::ylab(label="") + ggplot2::scale_y_reverse(breaks=seq(1,0,-0.25), labels = seq(0,1,0.25)) + ggplot2::xlab(label="Seasonal Total (mm)") + ggplot2::scale_x_continuous(breaks=seq(by=50, to=1200, from=100), limits=c(), expand=c(0.0, 0)) data_book$add_graph(data_name="merge2", graph=last_graph, graph_name="last_graph") data_book$get_graphs(graph_name="last_graph", data_name="merge2") rm(list=c("merge2", "last_graph"))