review | Thought splinters

Introduction to R for Social Scientists (I2RSS)

Sat, 02 Apr 2022 00:00:00 +0000

Acclamations for I2RSS

From my training (Ph.D. in sociology) but also from my different professorship on educational sciences, I was pleased finally to see a new R introduction especially directed at social scientists. The book title “Introduction to R for Social Scientists: A Tidy Programming Approach” (abbreviated I2RSS) was promising for me for two reasons.

The book is written by Ryan Kennedy and Philip Waggoner, two prestigious scientists involved in political and social research. Whereas Ryan has an astonishing career in political science, is Philip exceptionally experienced at an academic level in computational social science and data science. Philip is also a member of easystats (a software group that tends to an ecosystem of packages making statistics in R easy).

Promoting the modern tidyverse approach

The book features the tidyverse approach a modern collection of R packages that share an underlying design philosophy, grammar, and data structures.

Real data, results discussed and gentle dialog with reader

Without doubts, the book has several other merits: It uses real data sets, e.g., from the American National Election Studies (ANES) and teaches data management, visualizations, exploratory data analysis (EDA), and statistical modeling not in an abstract way, but also discusses the exciting results with their data sets. It talks to the reader in a very sympathetic and comforting way to master the (still very) steep learning curve for R beginners.

Some drawbacks and failures

But unfortunately, there are also two significant drawbacks to this introduction:

Educational inconsistencies

The book has several weaknesses from an educational point of view:

With some exceptions (data classes, data structures, and the map function from the purrr package is the chapter on “Essential programming,” in my personal view to advanced for R beginners, especially for people from the social sciences with no programming background.
Comparing the same results with base R and tidyverse code may be interesting for intermediate users. Still, it is a waste of time and complicates things for beginners. It is also misusing the sparse place in the book, which has only 198 pages.
The book uses R scripts and does not feature Rmd files. This is strange as literary programming is an essential step for reproducibility, and RMarkdown also facilitates interactive use of the program code.
Sometimes, the results of the code chunks are missings (e.g., graphs, tables, or reports). The explaining text under the code is, therefore, abstract. There are also some link errors, providing an URL starting with “www” where the referring webpage does not need/accept this addition.

These above weaknesses result in the deduction of one of the five stars.

Not a role model for Open Science

The second star deduction is motivated by missing teaching modern scientific behavior for early career researchers (ECRs). This shortcoming includes the prominent use of R Markdown and the use of collaboration and Open Science. I think that after the publication of Happy Git, there is no excuse anymore to teach Git and GitHub (or similar systems) even for R beginners. Another big mistake is not using the open-access politics of the bookdown.org infrastructure for the online version of the book.

In the book, the authors promote the friendliness and importance of the R community several times. They even mention that the book was initially written with the help of the bookdown package (page 2). So why not supply a role model for R beginners and young researchers?

Grammarly improves your writing

Fri, 07 Jun 2019 00:00:00 +0000

My story with Grammarly

This article reports on my experiences with the premium version of Grammarly, an AI-powered web service for grammar and spell checking. As this is going to be a very positive review, I feel the obligation to publicly disclose that I am not in any way affiliated with Grammarly Inc., the enterprise behind this product.

It was almost exactly one year ago in June 2018 when I learned the first time about Grammarly. I cannot remember anymore, why I stumbled over Grammarly. At first, I used it only with the free plan experimentally for some of my English writing on the web. I have to confess that I was very skeptical about the practical usefulness of this kind of app.

But after some tests, I was surprised about the quality of the suggestions. Grammarly was quite helpful as it gave me not only corrections for words spelled wrongly but also tips and ideas for a better style in English writing. At the same time, Grammarly told me not only how many improvements it has already suggested to me for a specific text passage, but also how many more suggestions would be in the premium version. Clearly enough, this was a sales gimmick, but a well-done one! I became curious about these other recommendations and signed finally up in August 2018 for the Premium version. At that time, it was still US$ 62,98 for a year, but in the meantime, the price has more than doubled (US$ 139,95) for the yearly subscription. I didn’t regret it!

Use cases illustrated by screenshots

Besides spell and grammar, Grammarly also helped me with style issues and even to find the right tone to address a particular audience. As a non-native speaker, it is difficult for me to judge any English improvements. But from reading much English literature, I’ve got the feeling that my English writing has improved tremendously.

I do not intend to write a tutorial for Grammarly, but I will give you some examples of use cases, illustrated by screenshots.

Use case 1: An already published article

I took a lengthy article (about 2.500 words), which I had published two years ago. I copied the markdown formatted text directly into the web interface of Grammarly. This article has a somewhat philosophical flavor so that technical concepts, which Grammarly has not in its dictionary, didn’t play a role. I had the text already cleaned up form spelling errors using RStudio’s spell checker. Nevertheless, Grammarly came up with more than 125 suggestions for improvements. I didn’t think that my English is so bad. How embarrassing!

When I inspected the text in detail, it turned out, that 19 suggested “improvements” weren’t useful, as they addressed quotes from English books by famous people like Michael Polanyi or Ludwig Wittgenstein. Another four recommendations related to markdown formatted text, where corrections would have resulted in syntax errors. But still there remained about 100 helpful tips! Statistically, Grammarly advised me about possible improvements every 25 words. I was shocked, and even a little depressed!

Grammarly suggests and explains a possible improvement

You see on the right top of the screenshot the number of alerts or warnings (= 120). Grammarly suggests finding a more common synonym for “emphatically” and explains with examples why it is maybe not the best choice. As all recommendations didn’t quite fit for my writing intention, I decided to delete this “overly complex” word without a substitution.

Use case 2: German text translated via Google Translate

I transcribed a German interview with Michael Rundel and ran the text through Google Translate. I believe, Google Translate has already overcome its start-up difficulties. Nowadays, I think, its translation is pretty good. But the writing must still be reviewed by a human.

As you can see in the image below, Grammarly recommended after an unpolished Google Translation from German to English a slightly better rate of improvements per word: 1892 words (bottom-left) with 104 tips (top-right) which accounts for one suggestion every 18 words.

Grammarly alerts in a machine-translated text

If you subtract the high number of 37 tips for punctuation issues (third line in the alerts list) – which partly are markdown formatting requirements and partly are differences between American and English punctuation rules – then the ratio improves to 1 tip for every 28 words. Admittedly this calculation is done under the caveat that after Google Translate, one has more sentences to restructure or to reformulate completely. But you can use Grammarly to implement these changes fairly quickly.

This calculation startles. Instead of struggling to formulate an English text from scratch and then to polish it with Grammarly, it appears to be more efficient – at least for me – to use another workflow: Writing my thoughts in a German paper -> than to translate it with Google Translate -> and finally using Grammarly to polish it.

Write the text in German

Translate it with Google Translate

Polish it with Grammarly

Publish it with Bookdown /Blogdown

But this reasoning is not entirely correct: The first version of the English text could already be done with Grammarly and would save one complete phase of the mentioned workflows. At least, I believe, this is valid for technical writing, where writing in English is not so complicated grammatically and therefore faster. But for publications in the Humanities, the workflow with Google Translate would be perhaps the better choice.

Use case 3: Writing English directly with Grammarly

To write text “live” in Grammarly has several advantages:

Grammarly gives you immediately a short explanation for every alert so that you can reformulate or restructure your document instantly.

Grammarly explains every current alert briefly and succinctly

Grammarly offers on-demand also longer clarifications and examples. So you can learn better writing during writing.

If required you can get from Grammarly a more profound explication with examples

Grammarly tutors you about the English language. This educational approach comes in handy when you notice that Grammarly often complains in similar situations. For instance, I didn’t know that there are different punctuation rules in American and British English. As I had written the lengthy text outside of Grammarly, I had to go through the whole article and change all occurrences of wrong punctuations one by one. If I had known these different rules, I would have to use the correct rules already in writing the draft version.

Grammarly offers on some subjects short tutorials

Grammarly supports different kinds of workflows: In the default configuration, Grammarly jumps automatically from one suggestion to the next. It is a good strategy for papers already written in English. But if you are writing directly within Grammarly, it is better to turn off this feature as it destroys the writing flow. Then you can select the recommendation you want to look at individually, and Grammarly jumps to the related text passage. A third option is to work on all tips concerning a specific category (e.g., spelling, grammar, punctuation, clarity, etc.) As you don’t have to jump from one type of problem to another one, the focus of one category of alerts speeds up the correction.

Working with Grammarly on alerts type ‘Punctuation’

Grammarly opens a window with synonyms when you double click on a word in your text.

Grammarly window to choose from a suggested synonym

Grammarly also offers recommendations to improve your writing style. The following screenshot presents my first try to begin this article. Grammarly analyzed it as a tedious text passage. This example demonstrates that Grammarly is far more than just a garden-variety of a spell or grammatic checker.

Screenshot of a Grammarly window with alert about a boring text passage

Grammarly varies its suggestions according to your writing goals. You can choose your intention, subject domain, target audience, your writing intention, your style (formal/informal), and even about the emotional flavor of your writing. Some of these features are still experimental, and I have to confess that I did not know when to choose what. I assume that my English is not good enough to use these advanced configurations.

Grammarly varies its alerts according to your chosen writing style

Grammarly can seamlessly be integrated via a browser plugin into types of web content (e.g., fora, web editing). But it seems to me that these possibilities are sparse. Yes, there is an add-in for Microsoft Word, but only for Windows. In WordPress, Grammarly does not work with the visual editor, you have instead to use the code editor. Likewise, you can’t use Grammarly with markdown files in RStudio or other text editors. But these shortages are not limitations as you can always write your book or article in the Grammarly cloud and then copy the text into your editor of choice. Grammarly stores your document in short time intervals, so you don’t have to worry to loose parts of your work. You can go back to unfinished articles at a later time and finish it for publication.

My personal desktop in Grammarly

I prefer to work with the browser extension, as it always presents the same user interface. Besides, it unleashes the full power of Grammarly. All screenshots in this post show the Grammarly editor in the browser, Google Chrome, in my case. But there are also plugins for Firefox, Safari, and Edge. All these browser extensions are free, and you can use Grammarly with limited functionality at no charge as well.

Other similar products

Finally, I should add that during the research of this article, I noticed that there are a couple of web services with related functionality available. The high number of recent articles about grammar apps insinuates a big market for this kind of service:

Wrapping up

All in all, I’m delighted with Grammarly. In my opinion, it is a beneficial tool for the non-native writer. It improves your writing for the web (e.g., posting in blogs or commenting in fora) but also for printed papers and books. It has many cute and even some outstanding features. Even with Grammarly, I am still not wholly comfortable to write in English, but at least I am stressed about it anymore.

I do not claim that Grammarly is the best product, as I do not know the other apps and have therefore no comparison. But in most of the reviews mentioned above, Grammarly is top ranked. It is undoubtedly the app with the biggest user base (over 10 Mio.).

Last, but not least, I would like to add that Grammarly is not idiot-proof. You have to make your choices, looking up other dictionaries to find better wording. (I am happy using all the time dict.cc). You have to be aware that even if Grammarly is AI driven, it does not understand your writing in the human sense. If you are writing a dumb and meaningless article, Grammarly cannot improve your reasoning and arguments. I noticed, for instance, that Grammarly sometimes had overlooked a missing word, an error a human would immediately see. And sometimes Grammarly is wrong with its suggestions as the last screenshot demonstrates. But these rare cases are easy to judge and correct.

Grammarly is not idiot-proof. Your own intelligent judgement is still required.

Alternatives for Disqus?

Thu, 14 Sep 2017 00:00:00 +0000

Disqus is a top-rated service for hosting and managing comments. But it has as an external service several disadvantages which opposed the philosophy of static websites diametrically. I discuss some alternatives for integrating discussion fora with static websites.

Disqus has a long list of drawbacks

In one of my tutorials I have explained how to integrate Disqus into your static website. But in this article, I recommend looking for alternative comment services. Why this different standpoint? In the meanwhile, I read several articles questioning Disqus. Here is a list of critiques I found on the web (without ranking):

Disqus is slow and has a bad (re-)loading behavior.
Disqus is tracking many different things for different customers, some of them hidden and unknown.
Disqus does not allow Markdown.
Disqus does not allow anonymous content (IP address, email, and name are recorded).
Disqus is hosted in the USA, which is considered to have less strict privacy laws than Europe.
Disqus is not open source.
Disqus does not allow that users can use free licenses for their comments. It is not clear who has ownership of the comments.
Disqus is one central authority collecting all comments of your website visitors and users on their servers.

A typical request log with Disqus enabled: There are 105 network requests vs. 16 without Disqus. Load-time rises between 2 to 6 seconds.

Especially the first and last bullet points defeat the advantages of a static website: Speed and all content always at your disposal. On a static website, data are just plain text files, saved locally on your hard disk. Therefore you can static sites transfer easily: Compress all your data in one zip file and unzip it where ever you want it. With Disqus, these advantages are not valid anymore because your blog text and its comments are hosted separately on different servers.

What are alternatives for Disqus?

Summarizing the disadvantages, I mentioned above I am looking for a system which is

free
open source
and does not host the user-generated comments centrally

The Hugo page on comments lists six available alternatives for comment on static websites¹. On AlternativeTo you will find 16 systems².

Cursory review on some services

Staticman

Screenshot of Staticman’s website

Staticman is open source and transforms user-generated content into data files to merge in your GitHub repository, along with the rest of your content. This approach seems promising for me, but until now, I couldn’t manage to install it. I have reported my problem, and I am currently waiting for help. As soon as I know how to fix it, I will review the system here on these pages.³

hyothes.is

Screenshot of hypothe.is' website

hypothes.is wants to provide “a conversation layer over the entire web that works everywhere, without needing implementation by any underlying site”. As far as I understand, for this approach, the data has to be stored centrally. So this software again is not an alternative to Disqus. This remark is not a critique of hypothes.is because it belongs to another category of software services. It has a new approach worth reviewing later in more detail.

IntenseDebate

Screenshot of IntenseDebate’s website

IntenseDebate is a feature-rich comment system for many blogging resp. CMS platforms. IntenseDebate is developed by the people who are behind many other well-known software services (e.g., WordPress.com, WooCommerce, Jetpack, Simplenote, VaultPress, Akismet, Gravatar, to name a few). It seems a bit odd that I could not find newer information on their blog than January 2014⁴. It is free but not open source, but it hosts the content centrally in the US. IntenseDebate is, therefore, no candidate for replacing Disqus.

Graphcomment

Screenshot of Graphcomment’s website

Graphcomment is a beautifully designed commenting service with a (limited) free plan. But it is disqualified under my criteria as the code is not open source and it hosts the comments centrally too.

MUUT

Screenshot of MUUT’s website

MUUT does not allow pre-moderation, e.g. every comment is online immediately. This is an interesting approach, but it is no alternative to Disqus: There is no free plan, and all rights of the user-generated content belong to MUUT!

ISSO

Screenshot of ISSO’s website

ISSO is a lightweight commenting service, programmed in Python, which allows anonymous comments. It is free, open-source, and installed locally. So it does qualify! But the installation procedure seems complex as there is no GUI, and one has to use the terminal for the installation. Furthermore, it seems to me that not all operating systems are covered. But I should give it a try anyway and review it.

Discourse

Screenshot of Discourse’s website

Discourse is a feature-rich open-source environment, supports Markdown and allows anonymous posting. As a hosting service, it has no free plan and costs a minimum of US$ 100 / month (with ~~80%~~ 85% discount for educational resp. 50% for non-profit institutions.). But you can install Discourse yourself without cost on your server. ~~Alternatively, you can pay a one-time fee of US$99 for a cloud installation with a US$10/month hosting fee.~~⁵ With the possibility to install it on your server, Discourse is another candidate to try out.

Using GitHub ()

Screenshot of Gazoo.vrv, Don Williamson’s website

Using GitHub: Another website is also recommending to use GitHub for comments. I have no clear idea how these proposed code lines will work in practice, but if it works, then it will certainly qualify: Open source, free, supporting Markdown and hosted by the website owner.⁶

Summary

I reviewed superficially different commenting systems. I was looking for a free, open source system, allowing anonymous content, Markdown and hosted by the website owner. Four services seem to fulfill my criteria: Staticman, ISSO, Discourse and a code proposal by Don Williamson. hypothes.is is an exciting project, but not a commenting system. It belongs, therefore, to a different category of software (annotation systems).

A competent review of these four systems requires a test installation, which I plan to do in the next few weeks.⁷

Update 2021-05-18: There are now ten services listed. ↩︎
Update 2021-05-18: There are 29 services listed, but not all are open source, and some have already discontinued their service. ↩︎
Update 2021-05-18: Even with some help by the developer I couldn’t manage to integrate it. Seems that my knowledge four years ago wasn’t enough. ↩︎
Update: 2021-05-18: Their blog is not updated anymore, but the main website has recent information, as the footer with copyright 2021 demonstrates. ↩︎
Update 2021-05-18: I adapted the changes of plans. ↩︎
Update: 2021-05-18: Now - four years later - I understand this approach perfectly. There are, in the meanwhile, already a bunch of different services using GitHub. In the new version of this blog, I am using utteranc.es a lightweight comments widget built on GitHub issues. ↩︎
Update 2021-05-18: I have never done these test installations. But at least I have revised this old article. I am still planning to look into details at least of some of these commenting systems. I am using now utteranc.es, and the Academic theme of this blog provides commento, another alternative to Disqus. ↩︎