Linear and Logistic Regression

drjwbaker commented 3 years ago

The Programming Historian has received the following tutorial on 'Linear and Logistic Regression' by @mjlavin80. This lesson is now under review and can be read at:

http://programminghistorian.github.io/ph-submissions/en/drafts/originals/linear-regression

http://programminghistorian.github.io/ph-submissions/en/drafts/originals/logistic-regression

Please feel free to use the line numbers provided on the preview if that helps with anchoring your comments, although you can structure your review as you see fit.

I will act as editor for the review process. My role is to solicit two reviews from the community and to manage the discussions, which should be held here on this forum. @svmelton and the English team have already read through the lesson and provided feedback, to which the author has responded.

Members of the wider community are also invited to offer constructive feedback which should post to this message thread, but they are asked to first read our Reviewer Guidelines (http://programminghistorian.org/reviewer-guidelines) and to adhere to our anti-harassment policy (below). We ask that all reviews stop after the second formal review has been submitted so that the author can focus on any revisions. I will make an announcement on this thread when that has occurred.

I will endeavor to keep the conversation open here on Github. If anyone feels the need to discuss anything privately, you are welcome to email me.

Our dedicated Ombudsperson is (Ian Milligan - http://programminghistorian.org/en/project-team). Please feel free to contact him at any time if you have concerns that you would like addressed by an impartial observer. Contacting the ombudsperson will have no impact on the outcome of any peer review.

Anti-Harassment Policy

This is a statement of the Programming Historian's principles and sets expectations for the tone and style of all correspondence between reviewers, authors, editors, and contributors to our public forums.

The Programming Historian is dedicated to providing an open scholarly environment that offers community participants the freedom to thoroughly scrutinize ideas, to ask questions, make suggestions, or to requests for clarification, but also provides a harassment-free space for all contributors to the project, regardless of gender, gender identity and expression, sexual orientation, disability, physical appearance, body size, race, age or religion, or technical experience. We do not tolerate harassment or ad hominem attacks of community participants in any form. Participants violating these rules may be expelled from the community at the discretion of the editorial board. Thank you for helping us to create a safe space.

Permission to Publish

@mjlavin80: please could you post the following statement to the Submission ticket.

I the author|translator hereby grant a non-exclusive license to ProgHist Ltd to allow The Programming Historian English|en français|en español to publish the tutorial in this ticket (including abstract, tables, figures, data, and supplemental material) under a CC-BY license.

anisa-hawes commented 2 years ago

Thank you, @svmelton! I'll work on the copyediting of these two lessons this week. I'll aim to complete by e.o.d Friday 22nd.

You didn't miss it, @drjwbaker. The revised editorial guidelines (that integrate copyediting as one of the final phases of our workflow) remain in draft, so aren't on the live site yet. They are currently being tested by the newest members of our EN team, to help ensure the guidance is clear and covers all that is needed.

anisa-hawes commented 2 years ago

Hello @mjlavin80,

I hope you're well.

I have applied copyedits to your lesson, which you can review in the commit History:

linear-regression.md edits here logistic-regression.md edits here

I've also made a list of the edits. This list includes some queries, which are marked with a Q.

linear-regression

l.36 “logistic and linear” > linear and logistic (to mirror the sequence of this two-part lesson).
l.38 Capitalise Assess, remove fullstop/period
l.40 Capitalise Interrogate, remove fullstop/period
l.44 Capitalise Explore, remove fullstop/period
l.47 Capitalise Analyze, remove fullstop/period
l.53-55 Suggest bulleted lists rather than numbered
l.53-55 Sentence-cap list items
l.59-65 Sentence-cap list items
l.71 add fullstop/period after etc.
l.73 It is discussed in two existing Programming Historian lessons. Create relative links, remove quotation marks
The first, uses
, while the second

l.74

picture a spreadsheet in which
create relative links, remove quotation marks
demonstrates how to create
matrices using

l.75

expand acronym TF-IDF upon first mention
create relative link, remove quotation marks

l.76

add link to Cultural Analytics article
create relative link, remove quotation marks
add authors’ names and add missing fullstop/period

l.81

remove quotation marks

l.85

and its partner, Logistic Regression with Scikit-Learn, will demonstrate
capitalise CSV (as it is written most consistently through the rest of this lesson)
create relative links

l.87

suggest split sentences to improve clarity,
For that article, I tagged each book review with a label representing the presumed gender of the book's author. I then trained a logistic regression model to classify the reviews as 'presumed male' or 'presumed female',
that reviews with both labels tends to use -> that reviews with each label tended to use.

l89.

several good reasons
for lessons (plural)
The way I will use this dataset for a logistic regression model

l.91

in this dataset
CSV uppercase
Q definition? Bag-of-words format

l.95-115, and throughout these two lessons, I have replaced bold (formatted with double underscore) with bold (formatted with double asterisk) for consistency across our journals

suggest bulleted list
suggest reduce additional spacing

l.130-334

suggest bulleted list
suggest sub bullets
suggest reduce additional spacing

l.116

differ from discrete variables, comma

l.120

whole numbers make sense but decimals don't
where both whole numbers and decimals are meaningful
scales, comma

l.125

split sentence: All this helps us to understand the difference between linear and logistic regression. Linear regression is used to predict a continuous, dependent variable from one or more independent variables, while logistic regression is used to predict a binary, nominal variable from one or more independent variables.

l.131

for a sample

l.133

What's important now, comma

l.141

average length

l.143

average length

l.145

add colon

l.150

a line-of-best-fit

l.152

The visualization above
add colon

l.159

add hard return after formula

l.165

add hard return after formula

l.172

We'll do something similar with logistic regression in the second of these two lessons.

l.184

in the first file,
In the second file, again

l.225 and 227

Q definition? I found myself Googling to define what an NA value is.

l.269

capitalise Python

l.271

create active link
add missing fullstop/period

l.273

add alert-warning box to replace Note
capitalise Windows operating systems

l.286

I noticed that endnote key [^16] was visible in the markdown….
there was an extra space at l.717, now deleted and solved!

l.295

Q definition? top K most frequent

l.332

words that do not match

l.346

Q definition? "K Best" see above, l.295

l.372

capitalisation of NumPy and SciPy (and throughout)

l.417

what it captures Q (or what is captured?)

l.496

for each column in our DataFrame
Here, comma

l.517

Q definition? Shapiro-Wilk test

l.519

Q definition? t-ratios
Q definition? p-values

l.524

This approach can become
this approach is

l.526

but this doesn't make much sense

l.532

to a point

l.559

use an F-Statistic for as a performance metrics -> using an F-Statistic as a performance metric

l.565

and then adding

l.643

On the other hand,

l.645

Q low coefficient scores will push it to an earlier date. ?

l.681

Q because unpublished books were more likely to be reviewed ?
add inverted commas: to the genre known as 'an appreciation'

logistic-regression

Suggestion: swap order “logistic and linear” > linear and logistic l.36, to mirror the sequence of this two-part lesson.
l.38 Capitalise Assess, remove fullstop/period
l.40 Capitalise Interrogate, remove fullstop/period
l.44 Capitalise Explore, remove fullstop/period
l.46 Capitalise Analyze, remove fullstop/period
l.52-54 Suggest bulleted lists rather than numbered
l.52-54 Sentence-cap list items
l.58-64 Sentence-cap list items

l.74

opening sentence is unclear to me: As with linear regression, logistic regression is easiest to describe when a problem with one continuous independent variable and one binary, nominal dependent variable is presented. Q How about:
As with linear regression, logistic regression is easiest to describe as an approach to modelling probability when one continuous independent variable and one binary, nominal dependent variable is presented.

l.74 -in this case scaled

l.76

suggest: find a way to avoid duplication of the phase “rooted in” in this sentence. How about:
Q are both based on the idea that conceptual binaries are rooted in linguistic conventions

l.76

that many variables which can be framed as binary for the purposes of logistic regression analysis, might otherwise […]

l.84

Q doesn’t merely ?

l.86

Q definition? sigmoid function
Q capital S for Sigmoid?

l.92

hard return after equation

l.94

add colon

l.96

remove fullstop/period

l.102

It's also important to understand

l.108

Q definition? to import pandas as pd

l.139

we run the

l.170

we used above
delete the sentence: “The top term list should be similar to the linear regression list, but it could differ since we're using a different sample of book reviews. “ which is repeated as the first sentence of the paragraph following this code block (para.203).

l.233

backticks around LogisticRegression

l.240

accounts for this difference

l.242

e.g., comma

l. 252

that our model has never seen before. We will evaluate

l.256

remove backticks from Scikit-learn. Add backticks to accuracy_score

l.263

or almost always, comma
looking directly at three

l.279

and were labeled m

l.285

has been falsely labeled.

l.287

how often our email client

l.289

it's not enough to know that
could be 99% accurate but fail

l.291

Q is actually also the same as recall ?

l. 293

it is calculated by
and then multiplying

l.315

lowercase f for Female
is fairly close to

l.319

the linear regression lesson

l.338

Q (for me to work out with @drjwbaker) I notice that the table isn’t displaying correctly in the Preview? All columns from and including “Correct” are truncated

l.394

that this feature will remain consistent

l.413

a list of the TF-IDF scores
we want to create

l.425

with the proportion

l.447

Q are there seven bins? (looks like 6? if there is one row per bin range as described l.425.)
Q “despite having made 11” where is this established?

l.474

we want to visualize

l.480

What's more, comma

l.500

, identical to the linear regression example,

l.538

share many terms in common with the results

l.576

version of my analysis.

l.617

to flag each step.
You'll notice new
The pandas merge() statement (singular)
The parameters left_index=True and right_index=True tell the method

l.625

add link

l.629

Here we encounter several

l.631

i.e., comma

l.637

who had become

l.639

The fact that this book's reviews
, and it is

l.693

First we have

l.697

together
Q some of the most important considerations that must be met ?
Once you fit these aspects

l.703

Q This is the first time you have capitalised Scikit-Learn — elsewhere through this lesson, you have written scikit-learn in lowercase. Which is your preference? (I have changed it to lowercase here, and in the remaining paragraphs). *For the title, I suggest using sentence-caps. This will be consistent with our other lesson titles, including those featuring Pandas (which is capitalised in lesson titles, but is written pandas lowercase in body text).

l.705

This is the first time you have capitalised Pandas. I’ve made it lowercase here.

l.715

documents’ (possessive)

Please let me know if you'd like to chat about any of these suggestions. Feel free to adjust my edits if there's something I've misunderstood, or if anything simply doesn't feel right to you.

I've prepped the Perma.cc links for both lessons, so they are ready to slot in as soon as these copyedits are approved.

Very best, and thank you for your patience. Anisa

--

Hello @drjwbaker,

Apologies for the delay. I hoped to complete this work on Friday, but it took me longer than I expected to work through both lessons.

I've made some minor adjustments to the YAML headers in both files:

I've sentence capped the lesson titles
Added hyphen and new line following editors:
Replaced mathjax: yes with mathjax: true (Or do these two do the same job?)
I have a question about the field output: pdf_document. I’m not familiar with this one?
One other technical note, is that I notice one of the tables in logistic-regression isn't displaying correctly in the Preview – all columns from and including “Correct” are truncated. It is the second table at Step 9: Model Validation or l.338-344 of the Markdown file.

Do you know what might be causing this? If not, Jenn may be able to help us with this (I think it is a 'table wrapping' issue).

That's all, I think. Thank you for your patience here.

A.

drjwbaker commented 2 years ago

@mjlavin80 : I see a small number of questions here. Do you have capacity to pick these up soon?

drjwbaker commented 2 years ago

I've made some minor adjustments to the YAML headers in both files:

Thanks @anisa-hawes!

I've sentence capped the lesson titles

Okay.

Added hyphen and new line following editors:

Okay.

Replaced mathjax: yes with mathjax: true (Or do these two do the same job?)

Don't know. I borrowed the sequence syntax from @acrymble's python articles. Hopefully this will work here?

I have a question about the field output: pdf_document. I’m not familiar with this one?

Me neither. I tend to just grab a recent header syntax and edit it. Feel free to add/remove any syntax you think we don't need (as you see these way more than I do!)

One other technical note, is that I notice one of the tables in logistic-regression isn't displaying correctly in the Preview – all columns from and including “Correct” are truncated. It is the second table at Step 9: Model Validation or l.338-344 of the Markdown file.

Hmm. No idea why this is happening I'm afraid.

anisa-hawes commented 2 years ago

Hello @jenniferisasi and @ZoeLeBlanc, Do you remember the table-wrapper fix for wide tables we applied to css/style.css in PR 2504? It seems like that functionality isn't quite working as it should do in the lesson logistic-regression.

This is a wide table, so our hope would be for it to be responsive/scroll-able, but in the Preview – all columns from and including the column headed “Correct” are truncated.

It is the second table at Step 9: Model Validation in the Preview, or at lines 338-344 of the Markdown file here. Might you have time to take a look? With many thanks, Anisa

mjlavin80 commented 2 years ago

@drjwbaker I think I can get to these by Monday or Tuesday. It's the last week classes here, so I'm a bit bogged down.

drjwbaker commented 2 years ago

Thanks so much @mjlavin80.

mjlavin80 commented 2 years ago

@drjwbaker @anisa-hawes I believe have completed responses to all of your questions. Should I push changes myself or just reply to this thread?

anisa-hawes commented 2 years ago

Many thanks, @mjlavin80!

Whichever is simplest for you!

Let me know if you'd like to discuss any of the edits or questions. I'd be pleased to connect and talk things through.

Very best, Anisa

mjlavin80 commented 2 years ago

Thanks @anisa-hawes I just pushed a round of changes. To summarize:

l.91

Q definition? Bag-of-words format (reworded to avoid)

l.225 and 227

Q definition? I found myself Googling to define what an NA value is. (added parenthetical)

l.295

Q definition? top K most frequent ... (added parenthetical)

l.346

Q definition? "K Best" see above, l.295 (hopefully above solution resolves)

l.417

what it captures Q (or what is captured?) (looks good as is)

l.517

Q definition? Shapiro-Wilk test (definition follows in original ... it is "a quantitative test to determine if a given sample is likely to have been drawn from a normal distribution")

l.519

Q definition? t-ratios
Q definition? p-values (reworded to drop both)

l.645

Q low coefficient scores will push it to an earlier date. ? (reworded)

l.681

Q because unpublished books were more likely to be reviewed ? (yes, books is right)

Logistic

l.74

opening sentence is unclear to me: As with linear regression, logistic regression is easiest to describe when a problem with one continuous independent variable and one binary, nominal dependent variable is presented. Q How about:
As with linear regression, logistic regression is easiest to describe as an approach to modelling probability when one continuous independent variable and one binary, nominal dependent variable is presented. (reworded in a different way)

l.76

suggest: find a way to avoid duplication of the phase “rooted in” in this sentence. How about:
Q are both based on the idea that conceptual binaries are rooted in linguistic conventions (sounds good)

l.84

Q doesn’t merely ? (yes, you're correct)

l.86

Q definition? sigmoid function (Changed first reference to "the sigmoid function" to "the logistic regression function" as the next sentence defines it.)
Q capital S for Sigmoid? (No, this is typically lowercase)

l.108

Q definition? to import pandas as pd (revised for clarity)

l.291

Q is actually also the same as recall ? (Seems unnecessary to me to add "actually also" here)

l.447

Q are there seven bins? (looks like 6? if there is one row per bin range as described l.425.) (I count seven (0-6 inclusive))
Q “despite having made 11” where is this established? (Changed to "As you may notice if you are running the code on your computer")

l.697

Q some of the most important considerations that must be met ? (yes)

l.703

Q This is the first time you have capitalised Scikit-Learn — elsewhere through this lesson, you have written scikit-learn in lowercase. Which is your preference? (I think your recommendation is correct)

anisa-hawes commented 2 years ago

Excellent. Huge thanks for your clarifications and replies to my queries here @mjlavin80 – I really appreciate your time.

My next step will be to replace external links in the lessons with archival perma.cc links (except in cases where the link directs to an action on the live web, e.g., install software).

I will also follow up with my colleagues to resolve an issue with the second table at Step 9: Model Validation or l.338-344 of the Markdown file, which I noticed isn't displaying correctly in the Preview.

drjwbaker commented 2 years ago

Great job all. @anisa-hawes: let me know if/when you need further input from me.

anisa-hawes commented 2 years ago

Just a note to say that I applied the perma.cc links to both lessons this afternoon.

I've also taken a look at css/style.css to double check the snippet which makes tables responsive on our site:

.table-wrapper {
  overflow-x: scroll;
}

It looks right... but I've noticed a wide table in another lesson on Submissions which is also truncated in this way, so something isn't quite as it should be. I'm following up with colleagues and will return with an update asap.

drjwbaker commented 2 years ago

@anisa-hawes Is this now just stuck with the table problem?

anisa-hawes commented 2 years ago

Hello @drjwbaker. Yes, that is the final thing to resolve. Spoke with Zoe again on Friday, and she assured me that this is on her list.

anisa-hawes commented 2 years ago

Dear @drjwbaker and @mjlavin80. Thank you for your extended patience. Good news! The table-wrapper issue is fixed!.

Checking the Preview here, (specifically the 2nd table at Step 9 Model Validation), I can confirm that the scroll action is working as it should. Thank you, @ZoeLeBlanc!

Hello @svmelton This pair of lessons are now ready for your final checks ahead of publication.

Key info via James's previous comment:

Linear Regression analysis with scikit-learn

Logistic Regression analysis with scikit-learn

  name: Matthew J. Lavin
  team: false
  orcid: 0000-0003-3867-9138
  bio:
      en: |
          Matthew J. Lavin is an Assistant Professor of Data Analytics specializing in Humanities Analytics at Denison University. His scholarship focuses on book history, cultural analytics, and turn-of-the-twentieth-century U.S. literature and culture.

drjwbaker commented 2 years ago

Woo! Apologies for the slow reply (covid wiped out last week for me). Anything else I need to do here @anisa-hawes?

anisa-hawes commented 2 years ago

Hello @drjwbaker and @mjlavin80. We are almost there! A few final things:

[x] @mjlavin80 could you send me an email (admin [at] programminghistorian.org) so that I can share our Authorial Copyright form for you to complete? This is a new element of our publishing workflow.
[x] @drjwbaker I just did a final check and see that we do have one last YAML field to populate: avatar-alt is missing for both lessons.

Sarah will take care of things from there, including requesting the DOI and slotting that in.

Something to think about for after publication is preparing a couple of posts for our Twitter Bot. We'll Tweet directly after we publish, but the Bot will help us to publicise the lessons in the future.

svmelton commented 2 years ago

Thanks @anisa-hawes! I've requested the DOI and will move forward with publication this week.

anisa-hawes commented 2 years ago

Hello @svmelton. A quick note to say that Matthew's bio has now been updated as part of PR #2611 so you don't need to do this when it comes to your final steps of the workflow.

[x] Hello @mjlavin80. I've emailed you our Authorial Copyright form. Let me know if you have any questions!
[x] Hello @svmelton. I've received Matthew's copyright form.

drjwbaker commented 2 years ago

Looks like these two articles we be publishes later this week. Huge thanks to @rcmapp and @thomjur for their thoughtful, constructive, and detailed reviews, to @mjlavin80 for two (as it turned out!) superb articles, and on the PH side to both @anisa-hawes & @svmelton for supporting the publication process. I'm really looking forward to seeing these articles appear!

rcmapp commented 2 years ago

Really excited to hear this! I'm happy to have contributed and will look forward to doing so again, whenever needed.

Best, Rennie

Rennie Mapp, PhD, MSMITProject manager for

strategic DH initiativesOffice of the VP for IT University of Virginia(717)713-1204

@. @.>*

On Wed, Jul 6, 2022 at 9:59 AM James Baker @.***> wrote:

Looks like these two articles we be publishes later this week. Huge thanks to @rcmapp https://github.com/rcmapp and @thomjur https://github.com/thomjur for their thoughtful, constructive, and detailed reviews, to @mjlavin80 https://github.com/mjlavin80 for two (as it turned out!) superb articles, and on the PH side to both @anisa-hawes https://github.com/anisa-hawes & @svmelton https://github.com/svmelton for supporting the publication process. I'm really looking forward to seeing these articles appear!

— Reply to this email directly, view it on GitHub https://github.com/programminghistorian/ph-submissions/issues/436#issuecomment-1176257839, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADA4MWXSETNI35XR66VREN3VSWGMDANCNFSM5I2JEZ6Q . You are receiving this because you were mentioned.Message ID: @.***>

drjwbaker commented 2 years ago

This feels stuck. Is there anything I can do to help?

anisa-hawes commented 2 years ago

Thank you, @drjwbaker. Sarah has prepared the PR to publish these lessons, but there is a problem with Jekyll's Build Checks, which I'm unable to fix without technical advice. I will update you here/in the PR as soon as a solution is identified.

A minor detail, but I did note that the two thumbnail images chosen to represent these lessons are colour rather than greyscale. For consistency, I think you could make them greyscale. That is not causing the hold up, but it is something to do in the meantime!

Thank you for your patience.

drjwbaker commented 2 years ago

A minor detail, but I did note that the two thumbnail images chosen to represent these lessons are colour rather than greyscale. For consistency, I think you could make them greyscale. That is not causing the hold up, but it is something to do in the meantime!

Any idea how I do that?

anisa-hawes commented 2 years ago

This is done. I have uploaded greyscale versions of both images to Sarah's branch. The Previews are currently available here and here, if you want to double check you are happy/see if you want to adjust the contrast slightly? (I think they look good).

svmelton commented 2 years ago

Hi all—happy to announce that this lesson has now been published. Congrats, @mjlavin80, and thanks @drjwbaker for your work editing this piece.

drjwbaker commented 2 years ago

We got there. Amazing. Thanks all.

rcmapp commented 2 years ago

Fabulous! So glad, and really appreciative of everyone’s hard work. 🥳

On Fri, Jul 29, 2022 at 9:33 AM Sarah Melton @.***> wrote:

Hi all—happy to announce that this lesson has now been published. Congrats, @mjlavin80 https://github.com/mjlavin80, and thanks @drjwbaker https://github.com/drjwbaker for your work editing this piece.

— Reply to this email directly, view it on GitHub https://github.com/programminghistorian/ph-submissions/issues/436#issuecomment-1199291685, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADA4MWR7654JIKN5SGOLITLVWPMSPANCNFSM5I2JEZ6Q . You are receiving this because you were mentioned.Message ID: @.***>

--

Rennie Mapp, PhD, MSMITProject manager for

strategic DH initiativesOffice of the VP for IT University of Virginia(717)713-1204

@. @.>*

thomjur commented 2 years ago

I am happy to see these two lessons out there, looking forward to using them in one of my classes! Great work everyone! <3

programminghistorian / ph-submissions