Hi,
I tried some plots and there seem to be problems communicating with the R-backend (I am using kubuntu edgy and upgraded the ubuntu package (rkward .36) to version .41).
Furthermore I was confused to find that there are no p-values for correlations - is there a way to display them?
Thirdly and generally, as I am SLOWLY working my way round R, I find it helpful to see the code that I generate using the GUI. In the ugly, but functional Rcmdr this is easier to understand - is there a way to keep the code displayed when using the rkward-menue as simple (and have all the extrainformation -for the nice table and all, I presume - processed hidden)?
best greetings
Matthias
p.s.: thanks for the answer to the last comment
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
> I tried some plots and there seem to be problems communicating with the R-backend (I am using kubuntu edgy and upgraded the ubuntu package (rkward .36) to version .41).
Not sure, what exactly you mean. Could you provide some more detail on what is / is not happening? I have simplified the scatterplot plugin somewhat in 0.4.2, which is to be released real soon, now, but it was working before, here.
> Furthermore I was confused to find that there are no p-values for correlations - is there a way to display them?
Not yet, unfortunately. In R, computing the correlation, and testing for significance is two different functions, which is why I did not add this right away, when I wrote the correlation plugin (long time ago). You can test for significance (pairwise) using cor.test (). Of course, sooner or later, this should be integrated into the correlations plugin.
> is there a way to keep the code displayed when using the rkward-menue as simple (and have all the extrainformation -for the nice table and all, I presume - processed hidden)?
No, not yet. As a short term amelioration, in 0.4.2, there will be "captions" on the code segments, to structure the generated code a bit. Also, the code for several of the older plugins (scatterplot and correlations among them) will have to be cleaned up considerably. Some are much more complex/messy than necessary, indeed.
In the past, the development focus has been on the IDE features of rkward. In number and quality of plugins (i.e. GUI access to statistics functions) we're still far behind RCommander (but see, e.g. the distributions plugins produced by sjar, for some development in that area). This will slowly improve over the next releases.
Regards
Thomas
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi,
I tried some plots and there seem to be problems communicating with the R-backend (I am using kubuntu edgy and upgraded the ubuntu package (rkward .36) to version .41).
Furthermore I was confused to find that there are no p-values for correlations - is there a way to display them?
Thirdly and generally, as I am SLOWLY working my way round R, I find it helpful to see the code that I generate using the GUI. In the ugly, but functional Rcmdr this is easier to understand - is there a way to keep the code displayed when using the rkward-menue as simple (and have all the extrainformation -for the nice table and all, I presume - processed hidden)?
best greetings
Matthias
p.s.: thanks for the answer to the last comment
Hi,
> I tried some plots and there seem to be problems communicating with the R-backend (I am using kubuntu edgy and upgraded the ubuntu package (rkward .36) to version .41).
Not sure, what exactly you mean. Could you provide some more detail on what is / is not happening? I have simplified the scatterplot plugin somewhat in 0.4.2, which is to be released real soon, now, but it was working before, here.
> Furthermore I was confused to find that there are no p-values for correlations - is there a way to display them?
Not yet, unfortunately. In R, computing the correlation, and testing for significance is two different functions, which is why I did not add this right away, when I wrote the correlation plugin (long time ago). You can test for significance (pairwise) using cor.test (). Of course, sooner or later, this should be integrated into the correlations plugin.
> is there a way to keep the code displayed when using the rkward-menue as simple (and have all the extrainformation -for the nice table and all, I presume - processed hidden)?
No, not yet. As a short term amelioration, in 0.4.2, there will be "captions" on the code segments, to structure the generated code a bit. Also, the code for several of the older plugins (scatterplot and correlations among them) will have to be cleaned up considerably. Some are much more complex/messy than necessary, indeed.
In the past, the development focus has been on the IDE features of rkward. In number and quality of plugins (i.e. GUI access to statistics functions) we're still far behind RCommander (but see, e.g. the distributions plugins produced by sjar, for some development in that area). This will slowly improve over the next releases.
Regards
Thomas