This is likely to change with time,
and will need to be regularly updated, but

Roughly thinking:

Pre-course/ initial lab: Installing R and R Studio
based on Getting used to R, RStudio, and R Markdown
Read: R Packages: A Beginner's Guide online

MD 1& 2 = Working with the RStudio IDE parts 1/2,
Introduction to R,
DataCamp Cheatsheets link,
Intermediate R
MD 3 = Introduction to the Tidyverse,
Data Vis w/ggplot part 1
MD4 = Introduction to Data,
Data Cleaning in R
MD5 = Data Manipulation in R with Dplyr,
Joining Data in R with Dplyr
MD6 = Exploratory Data Analysis,
Intro to Stats: Correlation and Linear Regression,
Correlation and Regression?? (seems like pick one of b or c)
MD7 = Intro to Stats: Mult Regression,
Multiple &Logistic Regression
MD8 = Foundations of Inference
MD9 = Inference for Numerical Data
MD10 = ??
MD 11= Inference for Linear Regression
MD12 = Communicating with the Tidyverse,
Reporting with R Markdown,
Building Web Applications in R with Shiny

Also, Introduction to Data (MD4?), Correlation and Regression (MD6?) might fit in somewhere
Maybe more case studies?

For example, in Ch4 Tidy:

life_expectancy <- read_csv('http://ismayc.github.io/le_mess.csv')

This book is a phenomenal resource! I'm using parts of it for a workshop I'm giving to graduate students in microbiology. The students have had no formal prior instruction in statistics (or have forgotten everything they learned in a stat 1000 class years ago).

Chapter 10 is an excellent introduction to the basics of hypothesis testing. For my purposes, the next logical thing to discuss after alpha and beta is power. I've pulled in relevant material from here: http://www.statisticsteacher.org/2017/09/15/what-is-power/ for the first iteration of my workshop, and suggest that a similar discussion would be at home in your book.

I tend to think that if I find something helpful others will too, so thought I would pass these thoughts along to you.

Again, thank you thank you thank you for developing this amazing resource.

"They were listed on the contexts of the box that the bowl came in"

Should be "contents", I think.

I have assigned my students to take the Data Camp class:

https://campus.datacamp.com/courses/effective-data-storytelling-using-the-tidyverse-free

After finishing Applying R Basics, they are confronted with an Upgrade Your Account to Continue screen.

With no obvious way to continue without paying.

I tested it myself and the same thing happened. I assume that the whole course should be free . . .

Are there Kindle and/or PDF versions of Modern Dive available for download? If not, there should be, as there is for OpenIntro Statistics. I realize that I could (probably!) make such a resource using tools provided via bookdown, but that is certainly too difficult for most users.

In Chapter 11, include links to all code/data referred to in current Section 1.1.1 What you will learn from this book.

What do we mean by data stories? We mean any analysis involving data that engages the reader in answering questions with careful visuals and thoughtful discussion, such as How strong is the relationship between per capita income and crime in Chicago neighborhoods? and How many f**ks does Quentin Tarantino give (as measured by the amount of swearing in his films)?. Further discussions on data stories can be found in this Think With Google article.

For other examples of data stories constructed by students like yourselves, look at the final projects for two courses that have previously used ModernDive:

• Middlebury College MATH 116 Introduction to Statistical and Data Sciences using student collected data.
• Pacific University SOC 301 Social Statistics using data from the fivethirtyeight R package.

Need to do a full pass over the book

Following the code in the "Modern Drive" markdown book located on bookdown.org I encounter the follow errors in the .md html output after running data(flights). Per Ismay book "Getting used to R, Studio, and R Markdown" there was comment about chunks--calling functions listed above will result in error if not called in the current chunk. Something to that effect. For the remainder of the chapter 3, I had to use the eval function to bypass the .md error in the console otherwise all processes halted. Any thoughts or instruction on where I went wrong?

When I run the comands of C.2.1 topic the graph remains empty!
Could you help me, or show the "right way" where I can make questions?
...
dyRangeSelector(dygraph(flights_summarized))

If use ts {stats} the graph appears but with X scale empty

Best Regards!

I think students may struggle a bit to understand when to use each of the three get_ functions since they are similar. Would be useful to add a closing summary to Chapter 7 discussing this. (Not pressing, but helpful to include on the next big release.)

Copied from corresponding Google Doc (only the first part):

1. Final edits to .Rmd files

a) Formatting

• Borders of all histograms should be white, making the binning structure easier to read.
• Use computer font for all computing concepts: function(), data_frame, variable_names, package_name.
• Remove all &, %, and _ in fig.cap for R chunk options since they break PDF build
• Ensure no code exceeds 80 characters. While HTML code block outputs tolerate >80, PDF code block outputs do not.
• Ensure skimr::skim() code is not actually run, but all calls and outputs are hard-coded (with hist preview removed and --- output cut down to 80 characters), since this will break all knitr::kable() code for rest of book.
• Apply {styler} code to all internal code (i.e. code the reader doesn't see) using styler's RStudio Add-In.

b) References

• Search for all broken references (search for @ref and ?? in html output)
• Maintain Chapter/Section/Subsection naming consistency
• Ensure all Figures & Tables are referenced.
• If possible, ensure index is up-to-date by adding \index{<term>} tags

c) Dataset management

• Remove all load() calls
• Ensure all CSV’s are loaded from moderndive.com and not other sites. See index.Rmd, set-options R chunk, copy all needed csv files to docs/
• Create bit.ly links for all linked Google Docs

d) Spell check

• Do it in RStudio.
• Scan over all changed content to make sure grammar is correct.

e) R Scripts: Make sure docs/scripts folder contains the appropriate scripts

• Development version should link to moderndive.netlify.app/scripts since code may be updated in development causing problems if someone on moderndive.com is trying to look over all the code for what is there
• If possible, go over all code blocks and ensure that purl = FALSE is set for all code chunks we don't want shared.

f) ETC

• Ensure only Shutterstock licensed photos/images are used.
• If possible, delete contents of rds/ and rebuild all .rds files in case any got stale.

2. Final sanity checks

• Update all R packages before one final build
• Build PDF version
• Ensure that .travis.yml does not have any GitHub development .9000 versions of packages.

3. Switch from dev to release version on GitHub:

a) Create new GitHub branch

• Create update-release branch to be used for pull request.

b) Edit index.Rmd

At the top

• YAML: Change r format(Sys.time(), '%B %d, %Y') to release date of form “January 1, 1970”

set-options R chunk

• Current version information:
  • Remove .9000 and bump version number
  • Set date to release date: replace format(Sys.time(), '%B %d, %Y') with date of form “January 1, 1970”
• Latest release information:
  • Update to release values
• CRAN packages needed:
  • Ensure only CRAN versions used
  • Periodically: Do a search for use of all packages and remove those no longer used

Lower in the file

• Comment out development version warning block
• Flip dev_version boolean to FALSE

c) Edit preface.Rmd

Section 1.4 “About this book”

• Latest published version: ensure info is correct
• Previous versions: Add previous version info
• Add previous version to previous_version/ folder. Steps:
  • Go to previous_versions/ and add new subfolder for soon-to-be previous version
  • Go to GitHub releases pages and download .zip of source code of soon-to-be previous version
  • In index.Rmd
    • Uncomment notice about this being an out-of-date version and hard-code version number
    • Ensure comment is surrounded by *** to highlight this note
  • Build book and delete redundant nested docs/previous_versions
  • Copy contents of docs/ folder to the new subfolder in previous_versions/

d) Edit other files

• NEWS.md: Update with all significant changes from this TODO
• .gitignore: Remove bib/packages.bib and docs/

e) Final step

• Clean and rebuild book

4. Publish release

• Merge dev-to-release-vX.X.X branch into master
• Merge master into release via PR (trick to remember: right into left). Or consider doing this?
• Rebuild release/docs and commit. You might need to remove docs/ from .gitignore to be able to commit to release branch
• Tag release on GitHub
• Add link to Previous versions around line 474 of index.Rmd of most recent previous version.
• Send email to MailChimp
• Make sure all links in left-hand navbar index work
• Relatedly, ensure moderndive package on CRAN is updated around same time as ModernDive book version bump.

4. Revert back to dev version on GitHub:

a) Create new GitHub branch

• Create release-to-dev branch to be used for pull request. See example commit on GitHub to revert back to devel version by undoing all changes to:

b) Edit index.Rmd

At the top

• YAML: Change release date back to r format(Sys.time(), '%B %d, %Y')
• Return “development branch” version warning block and flip dev_version flag

set-options R chunk

• Current version information:
  • Add .9000 to version number
  • Set date to r format(Sys.time(), '%B %d, %Y')

c) Edit other files:

• NEWS.md: Add section for new dev version
• Add bib/packages.bib to .gitignore
• Revert .travis.yml so that infer and moderndive packages are github dev versions

For the five models:

• 1 numerical x: intercept + slope
• 1 categorical x: baseline + offsets
• interaction: baseline, slopes, offset to baseline, offset to slopes
• parallel slopes: intercept + offsets + single slope
• 2 numerical: intercept + slopes

While figure linking within the text is working, the same links for tables are not. They all render as "Table ??" where the "??" links to a non-existent header (I think). See image:

Issue was present throughout chapters 8 and 9, haven't investigated further. From browsing Rmd it seems like the references are correctly spelling the names of code chunks.

For some reason, perhaps historical, it is unusual for people to include 95% CI with correlations. Instead they just discuss whether 'significant' or not.
If the get_correlation function automatically gave you the 95% CI, it would help educate people to stop interpreting correlation coefficients as precise estimates, particularly with small N.

I've been assigning my students the suggested DataCamp chapters at the beginning of each ModernDive chapter to accompany MD. The experience went well with MD chapter 3, which suggested chapters 2 and 4 of David Robinson's "Introduction to the Tidyverse" DC course. Even though the suggestions had students skip chapters 1 and 3, most were fine and could understand what was going on without those chapters. Plus, I had them finish those DC chapters when they got to MD chapter 5 (since they're suggested there).

When students got to MD chapter 4 on tidy data, though, most panicked and struggled a ton (and many just gave up). The suggested DC resource is chapter 3 of @apreshill's "Working with data in the Tidyverse" course. Unlike David Robinson's course, which students were somehow able to skip around in with no problem, Alison's DC course is pedagogically designed to build on previous chapters, so students needed to have completed chapters 1 and 2 to work on 3 (at least that's what the few that went back and did 1–2 told me—I'm not 100% certain of that because I haven't actually taken her DC course 😬).

It might be helpful to not suggest specific chapters from DC courses, but whole courses instead. MD does this later on—chapter 10 suggests the whole courses on "Inference for Numerical Data" and "Inference for Categorical Data", for instance, instead of chapters.

To use "open in new tab" markdown functionality. Ex:

[Map](https://deskarati.com/2012/03/31/london-undergrounds-real-map/)
[Map](https://deskarati.com/2012/03/31/london-undergrounds-real-map/){target="_blank"}

Currently they read from Hostgator folder previous_versions

Just like infer

First of all thank you so much for your book. It made sampling and hypothesis testing so much more clear for me. Something I never really understood in med school.

One thing I tripped over a little bit is the section "10.9.8 Simulated data" in the chapter on hypothesis testing the difference of means. Especially the "tactile experiment" wasn't clear the first time I read it. I needed like 5-6 reads to fully understand the idea behind the experiment:

The next step is to put the two stacks of index cards together, creating a new set of 68 cards. If we assume that the two population means are equal, we are saying that there is no association between ratings and genre (romance vs action). We can use the index cards to create two new stacks for romance and action movies. First, we must shuffle all the cards thoroughly. After doing so, in this case with equal values of sample sizes, we split the deck in half.

The thing that wasn't clear for me was, that you build two new stacks for "action" and "romance" movies that do not consist entirely of one genre but is a mixture of both. Since the H0 hypothesis is that there is no difference between the ratings of the two genres that's a valid way to simulate the H0 hypothesis. But that wasn't that clear for me. Maybe a small sketch of the experiment or something would make this much more clear. Even a sentence like "Note that the new stack for action and romance movies does not consist entirely of the specific genre" or something along those lines would be awesome.

I hope my point is clear. I would love to assist clarifying this section

Thanks again for your awesome content 👍

Comparison: Let’s compare the 3 standard errors we computed above in Table ??:

• https://moderndive.com/8-sampling.html#using-different-shovels

We present a wide array of such scenarios in Table ??.

• https://moderndive.com/9-confidence-intervals.html

Will add more as I find them.

Replace all load(url("http://www.openintro.org/stat/data/evals.RData")) with library(moderndive)

When knitting, error occurred with the following message:

Error reading bibliography .\bib/packages.bib (line 94, column 1):
unexpected "}"
expecting space or "="
Error running filter pandoc-citeproc:
Filter returned error status 1

SessionInfo:

R version 3.4.3 (2017-11-30)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

Matrix products: default

locale:
[1] LC_COLLATE=Chinese (Simplified)_China.936  LC_CTYPE=Chinese (Simplified)_China.936   
[3] LC_MONETARY=Chinese (Simplified)_China.936 LC_NUMERIC=C                              
[5] LC_TIME=Chinese (Simplified)_China.936    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] RevoUtils_10.0.7     RevoUtilsMath_10.0.1

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.15    bookdown_0.7    digest_0.6.15   rprojroot_1.3-2 backports_1.1.2 magrittr_1.5   
 [7] evaluate_0.10.1 stringi_1.1.6   rmarkdown_1.9   tools_3.4.3     stringr_1.3.0   xfun_0.1       
[13] yaml_2.1.18     compiler_3.4.3  htmltools_0.3.6 knitr_1.20 

@smetzer180 found in 9.6 EXAMPLE: One proportion we say the tactile_shovel1 dataframe is included in moderndive package, but it isn't (we manually create it). At the very least in short term show code that manually creates this, in medium term add to moderndive package.

Noticed that Learning Check Solutions are appearing. For example, https://www.moderndive.org/2-getting-started.html#nycflights13 . Just a note to remember to turn that off before Albert gets to teaching again.

Hi I recently discovered your book. It's very helpful and I am thinking of using it with my students.

I wonder if you have issues in section 7 starting at 7.7.4.
https://ismayc.github.io/moderndiver-book/7-hypo.html#data-1

movie_example.pdf

Suggestion: when comparing histograms for bootstrap and sampling distributions, would be better if one could be superimposed on the other: or at least have them shown side by side. Currently have to scroll up and down to compare them.

Re this section:
You can manually specify which continent to use as baseline instead of the default choice of whichever comes first alphabetically, but we leave that to a more advanced course. (The forcats package is particularly nice for doing this and we encourage you to explore using it.)
It is great to learn about forcats: I seem to remember stumbling on it after a very frustrating experience of trying to reorder levels of a factor. While I can see that you don't want to digress too much from explaining about regression, I wonder if it would not be worth saying a bit more about R's default behaviour with factor levels, as I think many people get stuck on this when they come to analyse their own data.

In section 2.3 you have the sentence: "By the end of this book, you’ll be able to better understand whether these claims should be trusted or whether we should be weary."

While I'm sure studying statistics makes people weary on occasion, I believe the word you want here is "wary".

Cheers!

Having just started recently, I am working my way through this very interesting project.
The knitting hangs at this piece of code contained in 11-inference-for-regression.Rmd:

null_slope_distn %>%
visualize(obs_stat = slope_obs, direction = "greater")

It says slope_obs object not found. Just by looking with my untrained eyes, it seems that that data frame is defined at a later stage.

Hi,

After teaching this past quarter, I realized the formula given for the t-test needs clarification. (I know you guys obviously know the difference, but I found myself wishing for more narrative around the formula provided in the text.)

The formula given is:

$$T =\dfrac{ (\bar{x}_1 - \bar{x}_2) - (\mu_1 - \mu_2)}{ \sqrt{\dfrac{{s_1}^2}{n_1} + \dfrac{{s_2}^2}{n_2}} }$$

But what is not said is that this the formula for Welch's t-test, where var.equal = FALSE:

The modified degrees of freedom would need to be clarified here as you can't use the formula given and actually do the t-test by hand without that too (and when teaching, helps students understand why all of a sudden the df can have decimals):

The t-test formula for var.equal = TRUE (in all its pooled SE glory) is here:

If you want an easy copy/paste, the formulas are in this raw Rmd.

Again thanks for the awesome text- I really enjoyed teaching with it this past quarter!

Alison

The coggle mind map for data visualization link is broken. Third paragraph in section 4.9.1.

The line across here from the output is weird. Maybe we can ask for the {skimr} authors to tweak this? skimr::kable() doesn't seem to work for me when I try it in bookdown either?

@ismayc Is there any reason to keep docs/ synced on GitHub, since travis-ci renders the book to the gh-pages branch? I feel like we should only sync changes to .Rmd and .yml files and images/data files, otherwise it increases the chances of merge conflicts and makes for harder commit/branch comparisons.

New desiderata:

• Transfer ownership of repo from https://github.com/ismayc/moderndiver-book to ModernDive GitHub organization https://github.com/moderndive/
• From there have
  • main branch be source code for latest released version
  • devel branch for source code for the development version
• Deploy/host the different versions of ModernDive via netlify as follows:
  • Have release version at moderndive.com deployed via netlify instead of HostGator as per this tweet
  • Development version at http://moderndive.netlify.com
  • Past versions output/source code at moderndive.github.io/<old_version>
• Fix bookdown.org ModernDive cover issue due to https on HostGator

Originally stated action items: Use travis deployment instead of manual builds for both ModernDive.com and

• Could use the gh-pages branch of http://github.com/ismayc/moderndiver-book to publish the book to instead using travis
• This allows us to only have to change the Rmd files on GitHub.com and have the page automatically update
  • Currently this is done via travis on:
    http://github.com/ismayc/moderndiver-book to create a new repo at
    http://github.com/ismayc/moderndive that creates a book at
    http://ismayc.github.io/moderndive and also at
    http://moderndive.netlify.com
  • There’s no need to push to a new repo though (could just as easily push to gh-pages on moderndiver-book instead)
  • This is the process I did with http://ismayc.github.io/chemistr-book
• Should also create a moderndive-releases GitHub repo that is updated only when we go to the next release
  • Then this repo could build via travis in the gh-pages branch and we could use netlify.com to come to populate ModernDive.com (also could get https support for free with netlify)

The book recommends a Data Camp class and provides this link

https://campus.datacamp.com/courses/effective-data-storytelling-using-the-tidyverse/

Although other Data Camp courses work fine for me, this one does not. I always get this message:

Setting up your session took too long.
If the problem persists, please report an issue.

Should this be working for me? I would like to assign it to the students in my Harvard class next week.

Hello,

Just wanted to give you a heads up that figure 6.2 isn't rendering. I just see a big white box.

Filed per https://twitter.com/ModernDive/status/1073340286091386881. May I suggest elevating the statistical background section to a short chapter.

Currently there are no tidy, bayesian, R-based introduction to statistics books that I am aware of. As such, initial statistics must be taught from a more traditional book (e.g. https://www.amazon.com/Introductory-Statistics-R-Computing/dp/0387790535/). It is unlikely an instructor would start with one such book for the basic first section and then transition to a more modern approach for future topics (confidence, visualization, modeling, hypothesis testing, etc).

Modern Dive could help solve this by adding an introductory chapter that expands the Statistical Background appendix. It can probably be less than the equivalent chapters in other stats books since, as was said in the tweet, the concepts will hopefully be interspersed within other sections to allow learning as doing. (I would recommend reviewing the other sections to ensure they do cover using the mean, median, mode, quantiles, SD, variance, and several common distributions such as normal/guassian, bernoulli, beta, biomial, uniform, geometric, poisson, gamma, log normal, exponential, and general power-law distributions).

I would suggest the goal is not to teach students when and how to use these concepts (as hopefully the rest of the book takes care of that), but provide context so that when they see them in use they understand how they fit into statistics as a whole. (For example, https://blog.cloudera.com/blog/2015/12/common-probability-distributions-the-data-scientists-crib-sheet/ gives an interesting quick explanation of basic distributions and their relationships.)

To that end, it may even be beneficial to mention common traditional statistics (p-value, t-test, etc) in this section and then point to the Appendix where they are explained, not necessarily to give students an alternative to the primary approaches taught, but simply so they understand where these things they will hear commonly sit in the context of what they have learned.

And thank you for what is ultimately the go-to reference for a tidy approach to statistics. I think it's sorely needed and an excellent book with or without modification. I look forward to buying a hard copy as soon as they come off the presses!

Need to do a scan over the whole book to make sure all references are matching.

Hi,

I am trying to render the book in html_book using the following command

bookdown::render_book("index.Rmd", "bookdown::html_book")

I encountered the following error:
Error in split_chapters(output, page_builder, number_sections, split_by, :
The document must start with a first (#) or second level (##) heading
In addition: There were 50 or more warnings (use warnings() to see the first 50)

What am I doing wrong?

Herman

Comparing the text to this article http://www.ejwagenmakers.com/inpress/HoekstraEtAlPBR.pdf it seems that the interpretation of the confidence intervals is not accurate. Although the bootstrapping process makes clear the frequentist approach, the interpretation seems to forget it.
I'd be happy to help review it if that'd be helpful.

Can't get options(knitr.kable.NA = 'NA')or just omitting it to work which should print NA as default in a PDF.

We could start by converting the Your Turns into Learning Checks.

This will serve as future release checklist as well.

• Just this time: get package list consistent
• Check for broken @ and ?? references
• Ensure Chapter/Section/Subsection consistency. Ex: 3/3.1/3.1.2
• Hide all Learning Check solutions
• Take banner off of master branch docs deploy after we create a dev branch.
• Make changes in index.Rmd to update versions of development and released.
• Tag release on GitHub for new released version.
• Update past versions list in Section 1.4. See #23 and #31
• Change travis to look for changes on the dev branch and then publish to gh-pages (for Albert's class on Monday).
• Create a referral page from https://github.com/ismayc/moderndiver-book to https://github.com/moderndive/moderndive_book (Contents in the old repo should be removed prior to the next major release--likely this summer.)
• Change links to script files to be to https://moderndive.com/scripts/ for the released version of the book. Not that big of a deal but it would be confusing if we changed some of the code in the development version of the book and then someone was trying to run the ModernDive.com code instead. We should probably set a toggle for this. (Delayed until next release.)

Maybe consider using standardization in place of normalization.

In my experience, it is common for students (and researchers) to mistakenly believe that normalization implies they are making the variable/data normal when in fact they are simply transforming their variable/data to a standard scale.

Just as the standard error is a special type of standard deviation, standardization can be thought of as a special type of normalization. It is such an important type that it gets its own name.

Section 10.8.1:

What is commonly done in statistics is the process of normalization. What this entails is calculating the mean and standard deviation of a variable. Then you subtract the mean from each value of your variable and divide by the standard deviation. The most common normalization is known as the z-score

Example 1

bootstrap_distribution %>% 
  visualize(obs_stat = x_bar)

yields

Warning message:
`visualize()` shouldn't be used to plot p-value. Arguments `obs_stat`, `obs_stat_color`, `pvalue_fill`, and `direction` are deprecated. Use `shade_p_value()` instead. 

Example 2

bootstrap_distribution %>% 
  visualize(endpoints = percentile_ci, direction = "between")

yields

Warning message:
`visualize()` shouldn't be used to plot confidence interval. Arguments `endpoints`, `endpoints_color`, and `ci_fill` are deprecated. Use `shade_confidence_interval()` instead.

Error 1

The following

sampling_distribution %>%
  visualize(fill = "salmon")

yields

Error in vapply(theory_types, function(x) { : values must be length 1,
 but FUN(X[[1]]) result is length 0

and was replaced with

ggplot(sampling_distribution, aes(x = stat)) +
  geom_histogram(bins = 10, fill = "salmon", color = "white")