Update tables and figures about community in the paper

[yenchiayi]

• [x] Table 1: top 3 packages in the detected community
• [x] Figure 4: naming convention within each community
• [x] Figure 5: features within each community typo: title of Figure 5

[yenchiayi]

Table 1:

[chainsawriot]

The last chart doesn't look correct to me. Need to check.

[yenchiayi]

I now use page_rank to rank the importance of packages in an identified community, instead of using evcent. Using evcent on a weighted directed graph may cause many problems, such as: https://lists.libreplanet.org/archive/html/igraph-help/2015-11/msg00020.html

After the modification, the top 3 packages of a community change slightly. I think the community names we came up with yesterday are still valid. Some are more clear, such as (comm_id: 25, comm_name: "Graph"), while some are a bit confused, as the similarity between id 42 and 44. I put my proposed revision in the column "suggested" below.

comm_id	n_mem	top	comm_name	suggested
6	5157	methods, stats, MASS	base
4	4758	testthat, knitr, rmarkdown	Rstudio
28	826	Rcpp, tinytest, pinp	Rcpp
3	463	survival, Formula, sandwich	Statistical Analysis
9	447	nnet, rpart, randomForest	Machine Learning
16	367	sp, rgdal, maptools	Geography 1	Map
15	131	gsl, expint, mnormt	Geography 2	Geography
25	103	graph, Rgraphviz, bnlearn	Bioconductor: Graph	Graph
49	79	tm, SnowballC, NLP	Text Analysis
42	55	tcltk, tkrplot, tcltk2	GUI	GUI 1
13	54	rsp, listenv, globals	Infrastructure 1
17	51	polynom, magic, numbers	Numerical Optimization
40	43	Biostrings, IRanges, S4Vectors	Bioconductor: Genomics	Genomics
77	38	RUnit, ADGofTest, fAsianOptions	RUnit
24	33	kinship2, CompQuadForm, coxme	Survival Analysis
2	32	slam, ROI, registry	Sparse Matrix
44	31	RGtk2, gWidgetstcltk, gWidgetsRGtk2	Infrastructure 2	GUI 2
75	29	limma, affy, marray	Bioinformatics
37	28	RJSONIO, Rook, base64	IO
45	27	rJava, xlsxjars, openNLP	rJava

Please kindly let me know your thoughts on whether to update the names. @chainsawriot @pymia

[yenchiayi]

The current revision is as follows. Please refer to the original files in the folder visualization_community/.

[chainsawriot]

@exilespacer I am okay with PageRank.

It seems to be clearer about what 15 is. It seems to be packages using GNU gsl library. These are more like statistical functions rather than geography.

[chainsawriot]

@exilespacer @pymia I suddenly have an alternative suggestion about the figures and tables in the paper. (Maybe in the presentation/poster, we can still use the assigned labels.)

How about we don't name the communities ourselves, instead simply use the top 3 packages such as "methods, stats, MASS" as the label? It has already implied that this community is about "base".

It can prevent the possible objection by the reviewers about our labeling (e.g. the geography stuff).

[pymia]

If it's possible to make the label like: Rstudio: testthat, knitr, rmarkdown ? For some communities like Base and Rstudio are easy to figure them out. But for other more domain specific labels, readers still need to try to identify which community that limma, affy and marray belong to.

[yenchiayi]

I agree with Mia's idea. My preference is as follows: LABEL: TOP 3 PKGS > LABLE > TOP 3 PKGS.

I actually used top3 packages as the community names before, and I think it is not effective in terms of information display. While it's true that we may label it not precisely, I found it hard to have a rough overview of the communities we identified, if only the top 3 packages are listed on the name. For example, if we label "studio," it immediately shows the popularity of the snack_name is likely because of the endorsement of the Rstudio community; if we don't, then we need further paragraphs in the main context to explain that. I would say better to make the figures and tables self-contained.

Though we find it hard to label, this difficulty would also apply to our readers, which I don't think they will google the description for each of the packages as we did and they may simply skip the section out of laziness to think. And even if they did, it is more efficient that we do it once and all the readers do not need to re-do the same things again and again.

I will suggest that we invite some people working with those packages to confirm if we make the label correctly. For example, I can post it on the FB pages of TW RUG and R-Ladies Taipei and collect some feedback. I knew there are some people from the field of Geographical data analysis and Genomics.

[yenchiayi]

I opened a new issue for the labels of identified communities in #26