M5 feedback

[yuliaUU]

Overall, I couldn't get a working app.R:

• libraries missing ( e.g.. ggplot2)
  running scripts produce error "Error in htmlBR() : could not find function "htmlBR"`
• another error error: could not find function "satisfaction_decision" ( as this function was somehow put in div_main ), have to be fixed

App Layout:

• I like the very simple layout! the color scheme also looks nice
• 1st graph: the majority of the points are so transparent that the graph appears almost blank. I would suggest revising the type of graph: I was not very sure how to interpret the graph. Maybe to use something similar to this

Inside the graph:

• Appropriateness of Mental Health option and some others have unedited y-labels: y-axis labels have number "7" while the rest of the options are text

slider:

• has the option "Prefer" but I have no idea what it means

suggestions:

• Maybe make the sidebar a bit wider? hard to read with only 4 words per line

[jacobgerlofs]

Hi @yuliaUU ,

Thanks for the feedback! I'm working on the first graph right now and have been thinking about visualizations. We were thinking of using a boxplot on top of the raw data points to help interpret, but wanted to check in to see if you also agree this would work well instead.

See picture:

Let me know what you think!

[yuliaUU]

What about a histogram? For each category make a count and then split it by other category as color

[yuliaUU]

Or may be radical barchart?

[jacobgerlofs]

Hi @yuliaUU ,

I think my main concern with these two plots is they don't highlight the relationship between the variables. The goal with the current plot is examine which variable(s) have the strongest relationship to one's satisfaction with their decision to pursue a PhD. I have considered doing a correlation for this reason too, but looking at the raw data points the variability and density of the data is quite heterogenous across levels of the variables and so I don't want to misrepresent the data with averages. I do think I could swing the original balloon plot you shared in your initial feedback in achieving this goal though!

[yuliaUU]

Well to answer your goal you have to do linear regression, as it is the only way to see the strength of relationship or boxplots that you currently have will do. Just looking at jitter points won't give any answer (in this case I think bar plot achieve the same result)