search

Sayani07 / paper-hakear

Algorithm for keeping and rating harmonies (HArmony KEeper And Rater)
0 stars 0 forks source link

Changes in Fig 7 & 8 - suggested by Di #1

Open Sayani07 opened 3 years ago

Sayani07 commented 3 years ago

The distributional difference do not look significant for the household ids chosen. A parallel co-ordinate plot with harmonies across x-axis and values of wpd across y, sorted by average value of wpd will help to identify most differing households for each harmony.

change 1: Add a parallel coordinate plot in Figure 7 change 2: Pick households which show more distributional difference.

Sayani07 commented 3 years ago

Fig 7 - Remove heatmap legend

Q. The original issue was that we don't have a limit for wpd, thus the legends are meaningless. Only colours are necessary, and it is preferable to delete the legend 

A. Legends for wpd is not removed.  patchwork is used to define the column widths and aggregate the plots, and when we choose guides="collect", all legends are shown. As a result, I was unable to delete the heatmap legend. However, we are already displaying the wpd scale for the third column, which leads me to believe that displaying the legend for the heatmap may not be an issue.

Sayani07 commented 3 years ago

Fig 7 - add parallel coordinate plot at the bottom

Q. Add parallel coordinate plot at the bottom of the heatmap and raw plot.

A. Placing it at the bottom makes the heatmap unreadable since the heatmap and raw plot are both compressed in length. Hence, I have flipped the coordinate and shown the parallel coordinate plot as a column. This

Sayani07 commented 3 years ago

Suggestion from Di: It might be good to colour the time series using the same colours as the par cords plot, to connect them, and sort from highest to lowest.

Sayani07 commented 3 years ago

Colours of the time series are made similar to the par cords plot.

Regarding sorting: The ranking of the time series will differ depending on the granularity of the comb. Did you mean sorting the average consumption level for the time series from highest to lowest? If so, could you please explain how it adds to the story or makes it easier to read? We're going to make the y-scales free anyway.