crtahlin
commented
10 years ago Opinion: 1.) As you've noted already, what could we do with graphs that have tables bellow them that additionally describe them? I do not like separating them, since then the user will have to switch tabs, which is not user friendly IMO. 2.) I do not like putting too many different things on the same tab (e.g. Tab 5). Although we could put check boxes on top, to enable user to filter what will be displayed, I think separate tabs are a much cleaner option. Otherwise the user will have to scroll quite a lot to see all the content. Also, checking of check boxes is another unnecessary step for the user. Since the graphs are quite different, they also "belong" onto different tabs contextually. 3.) Some things repeat in several tabs. This is both difficult to maintain (prone to bugs and headaches) and will probably be confusing to the user. 4.) Taking 1.&2.&3. into account, I would propose the following changes to the general structure:
- Summary = OK
- Graphical exploration -> Exploration (includes both graphs and appropriate tables)
- Regression analysis -> Analysis (more general, so we can add other stuff without reinventing things later)
The tabs would have names like:
- Summary: raw data
- ...
- Exploration: boxplots
- Exploration: profile plots
- ...
- Analysis: mixed models
- Analysis: RCS
- ...
Since we will be adding extra options to functions over time (e.g. non-linear mixed models) we cannot know exactly what will be required by the user in advance and it would not be prudent to name the tabs according to the options used on the plots (measurement occasion, time, ...). Even now, we already have the "days since inclusion in the study", not only "measurement occasion" and "time". If the user misses some options with, for example, mixed models, we can just add that to the tab containing them.
After some use, the user will know that the boxplots offer him visualizations of X, Y, Z kind, according to what he chooses.
What do you think?
llaarraa
As they are the tabs are still messy and difficult to describe in a general way in the paper. I would propose a general reorganization of the tabs - that should make the app more intuitive to use.
We should have apps of 3 types: summary, graphical exploration, regression analysis Summary tabs should contain tables. all the graphs are left to the graphical tabs?
Tab 0 - Welcome tab (?)
Tab 1 - Data summary: ok as it is, maybe move the missing value info in tab2 Tab 2 - Data summary: variables to analyze Provide an overall summary over time. Tab 3 - Data summary: variables to analyze by measurement occasion (tables with summary of data) Move here the tables now appearing under boxplots for numerical variables, produce something similar for binary variables. Add here all the information about the confidence intervals of the measures, the correction for multiple testing. What should we do about the graphs that refer to the data presented here? Display here the confidence intervals? Display here also the descriptive statistics graphically?
Tab 4 - Data summary: variables to analyze by measurement occasion and grouping variable Move here what is now appearing as table in Distr of var: by grouping variable Add here all the information about the confidence intervals of the measures, the correction for multiple testing. What should we do about the graphs that refer to the data presented here? Display here the confidence intervals? Display here also the descriptive statistics graphically?
Tab 5 -Graphical exploration: over time In a single tab we can include all the things that we have so far (and are appropriate for the type of data) let the user choose what to display and move the things in a single tab boxplots (numerical) profile plots (numerical) heatmaps (numerical and binary - to implement, see one of my previous posts) timeline (numerical, works in general? solved the problem? ok for binaries?)
Tab 6 - Graphical exploration: Clustering by measurement occasion (keep what we have) Should we add a clustering over time?
Tab 7? - Graphical exploration: by mesurement time Keep here the descriptive graphs? I would move all the graphs under data summary and remove this tab
Tab 8? - Graphical exploration: by mesurement time Keep here the descriptive graphs? I would move all the graphs under data summary and remove this tab
Tab 9 - Regression modeling: over time Similary as graphical exploration, we should have a single tab that does all time as measurement occ or days from inclusion estimates (and their CI) provided with tables and graphs ideally: we would add also the rcs option here -
Tab 10 - Regression modeling: by measurement occasion for binary variables we can include here the logistf and rcs options estimates (and their CI) provided with tables and graphs