rhoinc / web-codebook Goto Github PK

show top/bottom 5 if n > 10

suggestion from @agstn

At the very least add a note in the wiki about expected dimensions

Add a toggle that lets a user toggle between dot plots and bar charts for categorical variables.

Bar charts are simple enough without grouping variables, but when there is an active grouping (e.g. site), draw one panel per level like the SAS implementation here

continuous

Incorporate spike-histogram code in web-codebook instead of depending on the independent library.

... instead of rendering everything and then showing/hiding the current selection.

... especially if all of the levels have the same number of observations.

See comment here

0.10.0 lets you filter by group from none to any variable, but once you change the filter from none, it doesn't let you select it again unless you refresh the page

Add a process to either increase the bar width or (perhaps better) reduce the chart width.

Give the container a max width of 800 or 1000 pixels and change to use rectangles instead of lines.

Bars should be the same size, but they will be closer together and the overall chart width will be smaller.

Currently, makeBarChart() is used for variables with 0-5 levels and makeLevelChart() is used for 6+ levels. Right now this is just hard coded here and here and perhaps elsewhere too ...

When a categorical variables has more than 5 levels, render a bar chart showing all levels with y-axis = prevelance and x-axis = level. Hide the x-axis labels, and instead show the level name, number of observations and % of observations on mouseover.

This should do a better job showing the distribution for categorical variables with lots of levels (participant ID, etc).

In particular:
src/chart -> scr/codebook
src/chart/renderRow/makeChart -> src/charts

Add a "show all" button for truncated charts.

Descending prevalence, ascending prevalence or alphabetical

To reproduce:

Pull up CAT in FireFox: https://jwildfire.github.io/CAT/
Confirm the renderer is web-codebook renderer
Input version as: ded094e
Confirm data file is safetyData-queries/ADAE.csv
Click "Render Chart"
Click the grey arrow next to the AGE to expand the chart
Watch the chart expand for all eternity

Bring "Sex: M" etc off the edge.

Put the variable name and metadata above the chart instead of to the left so that the chart and the labels are all left justified and you take advantage of the horizontal space better.

If a user passes in an array of text (e.g. groups:["sex","race"]) then setDefaults should .map() it in to an appropriate array of objects.

Format legend items as "Level Name (n=XX)".

From @samussiah's suggestion here

The super-wide charts make comparison across levels difficult.

Charts are rendering outside of their wrapper.

:(

Default = 200ish? Start to see performance issues after that, especially when viewing grouped displays.

suggestion from @agstn

Up through v0.11.0 we used createChart to create Codebooks (createChart is used in almost all other renders also). v1.0.0 is going to move to createCodebook as the standard (since the object created is a codebook object containing lots of webcharts chart objects), but for convenience, let's support createChart() as an alias as well.

Remove the <title> and instead show tooltips on mouseover. Can probably re-use the histogram tooltip code.

See this comment

Consider a simplified view showing just "% true" for boolean variables. Would want to add an option that would allow a user to specify how boolean works in their data (0/1 vs. true/false vs. "yes"/"no").

When filtering on a variable (ex. age), the variable changed from categorical to continuous

Right now it's something like "0.1231098098" 👎

See code here

Came up in the SDF meeting on 4/13. Karen Boyle said it would be nice if we could select a sub-population and see that cohort on other charts in the Codebook, like JMP offers.

Feedback from the SDF meeting on 4/13. Karen Boyle said it would be helpful to see the x-axis of some of the charts shown in counts instead of percentages.

Make a panel with a searchable/sortable (choose an existing library) web table showing the raw data.

Use a vertical bar "|" instead of a circle to show the population mean in a grouped dot plot for a categorical variable.

Applies to:

• dot plots (custom legend)
• histograms (custom label on right-hand side)

This might be a good way to figure out how many bins are appropriate based on the data. https://en.wikipedia.org/wiki/Freedman%E2%80%93Diaconis_rule

For testing: number of bars will change when you filter. This would be one we may want to add a unit test to in the future. Goal is for it to look improved.

Potentially add margin and offset x-axis accordingly so that boxplot has more space below histogram. We will need to set this behind the scenes. We will test that we can change the size of the boxplot with difference configurations.

Fix: https://rhoinc.github.io/viz-library/examples/0009-web-codebook-demo/

Add examples for .makeExplorer()