Note: The trelliscope plot displays with blogdown would not have been possible without Yihui Xie’s lightning fast help. Not only did he implement the fix (in an afternoon), but documented/provided 2 options to fixing the trelliscope display issue. Thank you so much Yihui.

One of the impressive presentations at Jan 2017 RStudio Conference was Ryan Hafen’s trelliscopejs R package in action: Visualization in the Tidyverse. “TrelliscopeJS is a framework for creative interactive displays of small multiples”, and this is incredibly powerful when providing insights on product lines to decision makers.

User will be able to use the Sort, Filter, Labels, and Grid tabs to gain further insights from the data. A powerful way to dig deeper into data without having to produce several seperate plots. The display can be viewed full screen using the toggle @ right-hand corner.

Using nest() in tidyverse, rbokeh and trelliscope

Trellis_Data1 <- Product_df_1 %>%
  group_by(Products, Month, region) %>% 
  mutate(Unit.Cost = mean(Unit.Cost))

Trellis_Data2 <- Trellis_Data1 %>%
  group_by(Products, region) %>%
  mutate(industry_most = names(tail(sort(table(industry)), 1)))

 Trell_plot <- Trellis_Data2 %>%
  dplyr::group_by(Products, Month, region, industry_most) %>%
  dplyr::summarise(Qty = sum(Qty), Amt = sum(Amount), ASP = Amt/Qty) %>%
  group_by(Products, region) %>%
  nest() %>%
  mutate(panel = map_plot(data,
    ~ figure(toolbar = NULL, xlim = c(4.5, 12.5), ylim = c(21.5, 101)) %>%
      ly_points(Month, ASP, hover = .x[1:3], data = .x)
  ))

Trell_plot %>% trelliscope(name = "ASP_Month", nrow =2, ncol = 4, width = 500,  path = blogdown::dep_path())

The one variable I was not seeing on the display to sort and filter by was industry. Please see note from Ryan Hafen for an explanation as to why the variable cannot be used for display, and what can be done to address it.

Trellis_Data1 %>%
  group_by(Products, region) %>%
  summarise(uniq_industry = length(unique(industry)))

###Note from Ryan Hafen

To add additional variables to a display to sort and filter on, those variables just need to be present in the data. If it’s a numeric variable, a summary statistic (mean) will be computed. If it’s categorical, like industry, then the variable can only be included if it is constant within each panel (each panel can have a set of scalar metrics to sort and filter on - these cannot be vectors). The industry variable cannot be used in the above display because it is not unique within each combination of Products and region. You could create a new variable that is the most prominent industry for each Products / region combination and then it would be automatically included.

Trellis_Data2 <- Trellis_Data1 %>%
  group_by(Products, region) %>%
  mutate(industry_most = names(tail(sort(table(industry)), 1)))

Trellis_Data2

## # A tibble: 888 x 10
## # Groups:   Products, region [9]
##    Products Unit.Cost Amount   Qty Ship.Country       Year Month region   
##    <chr>        <dbl>  <dbl> <int> <chr>             <dbl> <dbl> <fct>    
##  1 ProdA         94.9  665       7 United States of…  2016  5.00 " North …
##  2 ProdA         94.9   96.0     1 United States of…  2016  5.00 " North …
##  3 ProdA         94.9  655       7 United States of…  2016  5.00 " North …
##  4 ProdA         93.9   93.9     1 Australia          2016  5.00 " Austra…
##  5 ProdA         94.9   93.9     1 Switzerland        2016  5.00 " EMEA " 
##  6 ProdA         94.9   96.0     1 United Kingdom     2016  5.00 " EMEA " 
##  7 ProdC         96.0  288       3 Germany            2016  5.00 " EMEA " 
##  8 ProdC         96.0  192       2 United States of…  2016  5.00 " North …
##  9 ProdC         96.0  192       2 United Kingdom     2016  5.00 " EMEA " 
## 10 ProdC         96.0  192       2 Ireland            2016  5.00 " EMEA " 
## # ... with 878 more rows, and 2 more variables: industry <fct>,
## #   industry_most <chr>

Using qplot and facet_treplliscope

#Does not work because of ggplot2
install.packages("ggplot2")
library(ggplot2)
qplot(Month, Unit.Cost, data = Trellis_Data2) + 
  xlim(4.5, 12.5) + ylim(21.5, 101) + 
facet_trelliscope(~ Products + region, nrow = 2, ncol = 4, width = 500, path = blogdown::dep_path("rmarkdown_files/Unitcost_vs_Month_gg"))

There will be future posts on gaining powerful multiple insights from displays of small multiples.

Thanks again for all your help Yihui & Ryan. Appreciate the time and effort you’ve spent in helping me resolve some of the issues I encountered while using the package.
More trelliscopejs.