I think the main issue with those tiny complex diagrams is coming from the fact that they had been designed for the full width of the page, but then were shrunk into a half page column. I really cannot understand why journals are still trying to save "paper" of the pdf documents compacting everything including small fonts.
I did not expect you to come out with something so extensive as of late, your previous videos covering the time it took for your PhD thesis as well as the coverage on how to not only write but pick out high quality IEEE journals in general were very insightful, so one hour going into depth on figures within research papers in particular (which I have not fully watched yet) is sounding very promising. Honestly we kind of should have seen this coming with how you covered the figures in the video talking about high quality IEEE journals, you did place quite the significance on them there so a video of this length and furthermore rigor makes sense. You're quite inspiring Electrical Engineering wise, keep up the great work by all means.
Great video! This sort of content isn't directly taught at college, even though it makes a huge difference on not only scientific papers, but also on every document we produce. Looking forward for the next videos!
I'd argue that vectorial images are not always better, they can and do occasionally look different in different pdf readers too. Some shapes might be disappearing or raster components of vectorial images are distorted even more than in fully rasterized images. I ruined one image in a paper just like that. The fact that some journal ruin the quality of raster images is kind of a separate problem and authors need to check proof, like you mentioned. In general, great video!
Thank you very much for this content, as a young researcher, I am just now realizing its importance. I will keep an eye out for the next video in the series.
Re: color, as an engineer, I think the rise of colors in plots has had a mixed impact. So I always assume: 1. The reader is colorblind, 2. the reader doesn't have a color printer, 3. the reader wants technical information. Obviously there are exceptions! But this is the default. Therefore, all graphics for a general audience should be colorblind friendly and grayscale friendly. For line plots, it's rare that I need to resort to color. If you look at any reference books or datasheets, the majority will be black and white, and not only because of limitations in cost of printing. Using linestyles, markers, and (often the best option) labels on the lines themselves or annotations, makes clear and concise figures (and legends are for markers only). There are exceptions of course - when lines overlap over a significant part of their length, markers alone cannot always save you. For other plots, like heat maps, 3D graphics, or bar charts, it's of course a different story. You can do a lot with labeled contours (and you should!), but it's not always enough. Just don't try to get too fancy with color schemes, a simple viridis or inferno is often sufficient - use the tried-and-true color schemes that have consistent changes in value! Rainbow is no good (as you say), single hue is great, and otherwise don't feel like you need to find something super aesthetic. What I hope a reader comes away from after looking at my figures is information. They shouldn't even realize they've seen a beautiful figure, it should be a natural experience. I love the impressive and well designed plots as much as the next person, and they're fun to make, but sometimes the real art is in restraint and pragmatism. IMHO Anyway, loved the video, enjoyed the advice and the practical examples are absolute gold, the good and bad!
This is definitely a nice video, and it covers a number of really important issues. On the issue of color, I think it's important to additionally stress that 1) Quite a large percentage of people are colorblind--around 5%. That's one in twenty! So, using color strategically, sparingly, and accessibly is key to maximizing your audience; 2) Contrast with a (usually white) background is also key. This video mentions yellow as problematic, but a lot of bright colors (cyan, e.g.) can be hard to distinguish when it's up against white. Another good reason to avoid bright colors (unless you can frame them in a dark color to help them pop). However, you also have to consider--is there contrast between the colors I'm using? This gets harder the more colors you need to use, so less is more; 3) As the video notes, you don't just want varied colors that play nicely together; you want varied saturation levels/hues too. If all the colors used are dark, for example, it may be hard to distinguish a dark purple from a dark blue from a dark red. So, using a color gradient that doesn't just vary color but also saturation (e.g., light red to dark blue) can increase contrast and accessibility, as those different hues will still be apparent in black and white.
I had been looking for a video like this for more than a year now. So glad to finally come across one! Keep up the great work, and thank you for helping to improve academia!
not only for the academic world but also when you work with data and want to make a visualization of it, this video is super good as it is never too much to learn how to make people understand you better through the data you present. very good video!
This comes at a perfect time for me. I'm currently developing a tool to help make more harmonious and accessible color schemes for figures in academia.
Thanks for the video, I'll forward it to the students in our department. I introduced a test for figures. It works like this: You print the figure (just the figure, in original size) and go to the collegue or student who has not worked on the topic and give him or her 30sec to look at the figure and ask for an explanation on what the figure show. You are not allowed to explain the figure yourself or help the collegue understanding. Afterwards, get back to your desk and update ;D
In my opinion you missed the most important point about colors. I am red-green colorblind, like 8% of men (XY) and 1% of women (XX). It is not that difficult to adapt color palettes to this, and many palettes are already created to be colorblind-friendly. Always keep it in mind please.
yeah I'm surprised he didn't mention grayscale/colorblind friendly palettes. Nice thing is a lot of tools plot colorblind friendly by default (except ggplot2 >:( ), but it's good to be aware of this stuff if you need to step out of the standard palettes. Like one figure I have is a diverging 5x5 color palette. That would have been MISERABLE if I didn't leverage existing palettes and use some color theory tools to get the gradations correct.
The one thing the video is missing (I know its outside the scope): How do I create the figures? Do you have a package you recommend? What package did you use for these?
Hi Jim! I plan to explain how I code and adjust my figures in the next tutorial. In short, I was using CorelDraw and plotting packages in Julia programming language.
Fantastic topic and interesting video, thank you! I find it to be one of the subjects which is unfortunately overlooked in most scientific trainings, while being extremely important. Most people start writing a new article by making figures first, and then building a story/text around it. This effectively makes figures the starting point of scientific writing. I spent countless hours reading about it during my PhD and later, especially color theory as you mentioned. I also find it to be one of the most satisfying things in research. It has a sort of artistic aspect, it's a bit like crafting visual art. I never thought about this eye/reading pattern, interesting! I would like to add a few personal remarks. First, even if you talked about it quite a bit, I believe that it is good to emphasize how symplifying figures is one of the most important aspects. I remember reading about the concept of minimizing "digital ink" as a principle to follow, and I find that it works extremely well to improve readability. I tend to remove as much clutter as possible, such as grids, lines, outlines (on the data points) etc. Regarding color, you forgot to mention the first constraint in this field imo: visual impairment. A significant portion of the population has some kind of visual impairment (often various types of color-blindness). This should be taken into account if we want as many people as possible to be able to read our research, and I make it the first criterion when picking a color scheme. Research on the topic helps picking the most universal palettes, I like Fabio Crameri's work for instance (he also provides some good palettes). One should also be careful about color palette uniformity (making sure that the relative distance between two colors is consistent across the scale) and meaning (as you mentioned in the video). I would advise, whenever it is possible, to combine color with other styling aspects to represent data. This can be line thickness, lightness, line style, point style, transparency... That way if two datasets are hard to distinguish by color, there are other hints available. Then as you mentioned, make sure that when the figure is visualized in shades of grey, it is still readable (most people end up printing articles in bw to read them). When it comes to integrating figures in the document, I mostly use Latex which outputs pdfs and deals quite well with figures in vector graphics. However the few times I had to use MSWord, I found such images difficult to integrate. I couldn't easily place a pdf or svg file in the text. I don't know if it is just me, but using the right tool for word processing with the right features is also quite important. Then journals often make a mess with manuscripts and figures, but unfortunately we cannot do much about it. Note that figure width should also be adapted whenever possible, some journals accept to break the two-columns style for figures. This can be helpful for wide figures, and there is an option in Latex to specify that a figure should take the whole width of the page. I suggest using it when necessary and possible (but to use it sparingly still). Finally my advice for those who work with programming/scripting languages to create figures (which you may cover in the next video), is to start by developing a template, and then to use it for every single figure in a project (article, report, thesis...). This helps immensely to obtain a consistent style. Every figure can then be tweaked as necessary, and it can be seen as some kind of safeguard against discrepancies. As a bonus, if you notice some styling error in all your graphs, it is likely that one slight modification in a single file will fix everything.
vector grasphics are nicer than raster but if you have to do it right you will tikz it. This way you ensure correct font settings and get optimal alignment in your document. Latex has problems arranging svgs. A large amount of svg will cause performance issues, consider using pdfs in this case.
I’m really excited about this video and the rest of the series! I’m about to start my PhD, and I find this topic quite challenging, especially since there’s so little quality information available. Most resources only cover basic stuff like bar graphs or using Excel, but this tutorial is exactly what I needed! Thanks so much for all the effort you put into it :)
So I will generally agree with "no complimentary colors", however there are ways to muted the colors so that they don't clash. I'm a fan of a hazy blue and a hazy, slightly darker orange. blue and orange are the most colorblind friendly complimentary, they are the easiest on the eyes, and with some darkening/greying, they can look quite nice.
Really great info here! Thank you! Would love a follow up about how to practical implement some of these ideas (i.e. combined vector-raster figures) in python and latex.
A few days ago I had to give a presentation about a research project. Thanks to your video I now noticed some mistakes I made, especially the pixelated images. 😀👍🏻
Can't fully agree with color advice... Complementary colors are great when there are few strokes or spots which must be easy distinguishable. But it's important that colors don't become monochrome gray blob when viewed, converted or printed in gray-scale, so it's better to chose colors of different brightness.
For colour palettes there are better tools that adjust for human perceptual colour difference. OKLAB pioneered it, if you want smooth gradients. LCH pickers are not bad. Some compensate for colour blindness too.
This is such valuable advice in general, should be mandatory in high school. You should review some of my published figures! Always open for public criticism haha
Great video, great idea. The most overlooked aspect of science. You make some very good points, but I think you lost it in the colors, a very hard and subjective matter, more often Red is not error, that is incorrect assumption. Although, red is usually the pen that a teacher use to correct and sometimes the color for the error bars, red more often means: fire, hot, much or the main point, it depends a lot on the subject and cultural experience. There is the classic blue to red for low to high values. In my opinion your plot should be black and transparent (gray) for the error figure. Also I believe nobody reads first the title of the figure, usually people read the plots like onions first the content and then the text that explains the content, layer by layer till the last detail. Please keep publishing more on the topic so we can all learn.
Hey Andrey, thank you for the most appreciated video, I have seen many videos in this regard but none of them was so detailed and pinpointed the actual problems I was confronted with. If I were to give a wish for your nexr video topic, to show us what software you use to create your unique graphs and diagrams? There are a lot of software like Excel, SPSS, Graphpad Prism, but what is your secret tip and how to get started, it would also fit alot to add the new A.I. tools that promise to solve your statistical problems and automatically generate charts and graphs.
Hi! Thanks for your feedback. Sure, I will show how I create figures. My main tool is the Plots package in Julia. I also use CorelDRAW to adjust and modify the figures before putting them in the paper. Hopefully, I will manage to make the next tutorial in November-December. Andrey
I believe most students underestimate how much thought must be involved in creating good figures, but it could also be my own training which was lacking. There are entire books on the subjects!
Hello Andrey! Thank you so much, everything is amazingly explained. Can you tell please, is saving the illustrations in tiff format good choice? Or better to save in svg or pdf? P.S I'm a physiologist and I use inkscape in making biological illustrations
Thanks Alex! I've never used tiff for my papers. As I see on Wikipedia, this is a raster graphics image format. This means that it will not save your results as scalable vector graphics. However, I don't know your images. Maybe tiff works well for your needs. Is tiff a popular format in your field? Andrey
@@chuscience Other thing I noticed when I saved my illustrations in PDF format. Even though I've drawn two lines which perfectly fits to each other, in PDF when you put the zoom on 75% one line appears longer than the other (they don't fit) but when you zoom in (100% and more) they fit to each other and appear in same length. It is so weird
Great video. How do you feel about ridgeline plots? I personally don't find them appealing, they always appear messy to me. Yet I see them everywhere. Second question, like almost 10% of all men I am red-green colourblind and so really struggle picking the right colours for my plots. Do you have any tips for me? Thanks!
Hi! I haven't used Ridgeline plots in my papers, however I used Violin plots a few times. I think they are somewhat similar. Indeed, if we have too complex and messy data to present (say, many strange overlapping distributions), such plots are not the best choice. But they can look great in other cases. So my advice would be to test them and see if your figure looks nice with such distribution plots. Think about if readers will be able to understand the message of the figure. I can't give great advice regarding picking colours for colourblind. Maybe select just one colour? Say, blue, and anything else is grey and black. Not the most intense blue though, maybe some shades of blue. So the main colour scheme would be a gradient from grey to blue, plus some black and white elements. Andrey
@@chuscience Thanks! One more thing I noticed is that in density plots with pixelated grids my brain tells my eyes I have to focus, to make the picture sharper. That makes it unpleasant to look at for longer periods. For example, the plots shown in the AlgebraOfGraphics tutorial have this effect on me. I wonder if I'm the only one or if this is a known thing?
Something I had a discussion about with colleagues in the past: When (if ever) is it okay to trace over pixelated graphs with vector software, to give a nicer looking plot? Sometimes, you cannot (easily/realistically) get raw data out of a piece of measurement equipment for various reasons (I've even had this for security reasons before - you were not allowed to take measurements out, only the pngs of the plots). In that case, I cannot get a nice vector graphic plot. Is it okay for me to trace the data in an vector graphics program? What about just drawing ledgends over it by hand, to make the graph more readable? Is it okay so long as I clearly state that it is a trace? I've also done the same when adapting data from old figures, in papers that are only available as low-resolution scans. What about those cases?
Hi Vaes! I think that tracing pixelated data is fine as long as you do it in an ethical way. That is, you trace all data and don't manually modify it to hide some problems or to highlight only the things you want to show. If your ultimate goal is to make data nice and clear for readers, then tracing is a good idea. For this purpose, manually adding legends and extra notes should also be acceptable. You can mention in a paper that the figure is traced and edited as a vector plot. For raw figures from the software, the readers can check your archive/repository [link]. Andrey
Could you consider in the next video on how did you import the figure into word... i think many students dont know how to import vector in the word file not only the issue of knowing them
If you are willing to put effort for nice figures, I would highly recommend using Latex to prepare your paper. Word might be easy to use but (to me) has many limitations and tripping stones. Yes Latex has a learning curve and once figured out I can guarantee you will have better documents just from the text alone. In technical documents consitancy is important and automation aids this immensely which is (almost) entirely unavailable in Word (aside from the build in features). Already creating equations in Word is a mess.
Hi! Great question. I don't think there is a single correct answer. You may need to test different fonts for your figures to see if they look nice in a manuscript. Some people advocate for Arial since it is simple and very well-known. Other options are Computer Modern (because of the LaTeX style), Georgia, Garamond, Courier, etc.
In the Windows system, PDF vector graphics cannot be directly inserted into Word and need to be converted to EMF format. The best way is to open it with Inkscape, making sure to use Cairo import instead of internal import, and then save it as .emf format. However, in the Mac OS system, PDF vector graphics can be directly inserted into Word with good results because Mac's Word converts PDFs into EMF format images. It is very strange that Microsoft's own Windows does not allow Word to smoothly insert PDF format vector graphics.
As someone who does structural analysis, I never understood why programs use rainbow stress plot results for finite element analysis. It makes more sense to use 2-color gradients for positive and negative stress respectively. Doing this allows a sense of relative scale without having to reference the legend for every color.
It’s called a paper. Not a playground. If you want that you find in computer graphics or some areas of AI complementary websites that 10-20 people or so have worked on in addition to the manuscript. If you want that, put in the effort and contribute such playgrounds to your field of research.
19:29 I bet if you make those numbers bigger then they will either obstruct the diagrams or they will be too bunched up together because there are a lot of numbers.
Yes, but this means the whole diagram type should be changed. Maybe it should be presented as a simplified graph with only branches and nodes, without generator symbols, substation buses, etc. Maybe no heatmap figure is needed here at all? What is the message here: that some locations have more capacity. Perhaps we should illustrate this distribution using some other plots. Or, if keeping the maps, they should be allocated much more space.
![The moment we stopped understanding AI [AlexNet]](/img/1.gif)
As a current undergraduate at Imperial EEE - This is insane.
As an aspiring academic, I hope we cross paths sometime in the next academic year !!
Hi, Amin! I'd be happy to catch up for a coffee. Please drop me an email when you are around the campus.
@@chuscience Very much appreciate the offer!! Will definitely reach out sometime during the term.
I think the main issue with those tiny complex diagrams is coming from the fact that they had been designed for the full width of the page, but then were shrunk into a half page column. I really cannot understand why journals are still trying to save "paper" of the pdf documents compacting everything including small fonts.
I did not expect you to come out with something so extensive as of late, your previous videos covering the time it took for your PhD thesis as well as the coverage on how to not only write but pick out high quality IEEE journals in general were very insightful, so one hour going into depth on figures within research papers in particular (which I have not fully watched yet) is sounding very promising.
Honestly we kind of should have seen this coming with how you covered the figures in the video talking about high quality IEEE journals, you did place quite the significance on them there so a video of this length and furthermore rigor makes sense. You're quite inspiring Electrical Engineering wise, keep up the great work by all means.
Great video! This sort of content isn't directly taught at college, even though it makes a huge difference on not only scientific papers, but also on every document we produce. Looking forward for the next videos!
This video was amazing! I absolutely loved it, and I can't wait for you to release the second part! Keep up the great work!
I'd argue that vectorial images are not always better, they can and do occasionally look different in different pdf readers too. Some shapes might be disappearing or raster components of vectorial images are distorted even more than in fully rasterized images. I ruined one image in a paper just like that. The fact that some journal ruin the quality of raster images is kind of a separate problem and authors need to check proof, like you mentioned. In general, great video!
Thank you very much for this content, as a young researcher, I am just now realizing its importance. I will keep an eye out for the next video in the series.
Great job Andrey. Your video tutorial is gold material. 😎👍
Re: color, as an engineer, I think the rise of colors in plots has had a mixed impact. So I always assume: 1. The reader is colorblind, 2. the reader doesn't have a color printer, 3. the reader wants technical information. Obviously there are exceptions! But this is the default. Therefore, all graphics for a general audience should be colorblind friendly and grayscale friendly.
For line plots, it's rare that I need to resort to color. If you look at any reference books or datasheets, the majority will be black and white, and not only because of limitations in cost of printing. Using linestyles, markers, and (often the best option) labels on the lines themselves or annotations, makes clear and concise figures (and legends are for markers only). There are exceptions of course - when lines overlap over a significant part of their length, markers alone cannot always save you.
For other plots, like heat maps, 3D graphics, or bar charts, it's of course a different story. You can do a lot with labeled contours (and you should!), but it's not always enough. Just don't try to get too fancy with color schemes, a simple viridis or inferno is often sufficient - use the tried-and-true color schemes that have consistent changes in value! Rainbow is no good (as you say), single hue is great, and otherwise don't feel like you need to find something super aesthetic.
What I hope a reader comes away from after looking at my figures is information. They shouldn't even realize they've seen a beautiful figure, it should be a natural experience. I love the impressive and well designed plots as much as the next person, and they're fun to make, but sometimes the real art is in restraint and pragmatism. IMHO
Anyway, loved the video, enjoyed the advice and the practical examples are absolute gold, the good and bad!
Thank you, your points were foundational missing in the presentation before.
This is definitely a nice video, and it covers a number of really important issues. On the issue of color, I think it's important to additionally stress that 1) Quite a large percentage of people are colorblind--around 5%. That's one in twenty! So, using color strategically, sparingly, and accessibly is key to maximizing your audience; 2) Contrast with a (usually white) background is also key. This video mentions yellow as problematic, but a lot of bright colors (cyan, e.g.) can be hard to distinguish when it's up against white. Another good reason to avoid bright colors (unless you can frame them in a dark color to help them pop). However, you also have to consider--is there contrast between the colors I'm using? This gets harder the more colors you need to use, so less is more; 3) As the video notes, you don't just want varied colors that play nicely together; you want varied saturation levels/hues too. If all the colors used are dark, for example, it may be hard to distinguish a dark purple from a dark blue from a dark red. So, using a color gradient that doesn't just vary color but also saturation (e.g., light red to dark blue) can increase contrast and accessibility, as those different hues will still be apparent in black and white.
I had been looking for a video like this for more than a year now. So glad to finally come across one! Keep up the great work, and thank you for helping to improve academia!
not only for the academic world but also when you work with data and want to make a visualization of it, this video is super good as it is never too much to learn how to make people understand you better through the data you present. very good video!
This comes at a perfect time for me. I'm currently developing a tool to help make more harmonious and accessible color schemes for figures in academia.
I am very thankful for your effort. You do a great job and many young researcher will learn a lot from you!
Your video has been very helpful, I look forward to the next ones.
Keep it up 😁
Thanks for the video, I'll forward it to the students in our department. I introduced a test for figures. It works like this: You print the figure (just the figure, in original size) and go to the collegue or student who has not worked on the topic and give him or her 30sec to look at the figure and ask for an explanation on what the figure show. You are not allowed to explain the figure yourself or help the collegue understanding. Afterwards, get back to your desk and update ;D
In my opinion you missed the most important point about colors. I am red-green colorblind, like 8% of men (XY) and 1% of women (XX). It is not that difficult to adapt color palettes to this, and many palettes are already created to be colorblind-friendly. Always keep it in mind please.
the best way to be sure is to make it grayscale-friendly aswell, so if you have a green and a red, have them in different brightnesses.
yeah I'm surprised he didn't mention grayscale/colorblind friendly palettes. Nice thing is a lot of tools plot colorblind friendly by default (except ggplot2 >:( ), but it's good to be aware of this stuff if you need to step out of the standard palettes. Like one figure I have is a diverging 5x5 color palette. That would have been MISERABLE if I didn't leverage existing palettes and use some color theory tools to get the gradations correct.
Amazing content, Andrey!
The one thing the video is missing (I know its outside the scope): How do I create the figures? Do you have a package you recommend? What package did you use for these?
Hi Jim! I plan to explain how I code and adjust my figures in the next tutorial. In short, I was using CorelDraw and plotting packages in Julia programming language.
@@chuscience thanks
@chuscience I ❤ julia
3:30
I personally use TikZ. It's very tricky to use and has a huge learning curve, but oh boy does it produce pretty plots.
Excellent content. I look forward to seeing some code snippets implementing the concepts discussed here.
high quality tutorial. many thanks
Fantastic topic and interesting video, thank you! I find it to be one of the subjects which is unfortunately overlooked in most scientific trainings, while being extremely important. Most people start writing a new article by making figures first, and then building a story/text around it. This effectively makes figures the starting point of scientific writing. I spent countless hours reading about it during my PhD and later, especially color theory as you mentioned. I also find it to be one of the most satisfying things in research. It has a sort of artistic aspect, it's a bit like crafting visual art.
I never thought about this eye/reading pattern, interesting! I would like to add a few personal remarks.
First, even if you talked about it quite a bit, I believe that it is good to emphasize how symplifying figures is one of the most important aspects. I remember reading about the concept of minimizing "digital ink" as a principle to follow, and I find that it works extremely well to improve readability. I tend to remove as much clutter as possible, such as grids, lines, outlines (on the data points) etc.
Regarding color, you forgot to mention the first constraint in this field imo: visual impairment. A significant portion of the population has some kind of visual impairment (often various types of color-blindness). This should be taken into account if we want as many people as possible to be able to read our research, and I make it the first criterion when picking a color scheme. Research on the topic helps picking the most universal palettes, I like Fabio Crameri's work for instance (he also provides some good palettes).
One should also be careful about color palette uniformity (making sure that the relative distance between two colors is consistent across the scale) and meaning (as you mentioned in the video). I would advise, whenever it is possible, to combine color with other styling aspects to represent data. This can be line thickness, lightness, line style, point style, transparency... That way if two datasets are hard to distinguish by color, there are other hints available. Then as you mentioned, make sure that when the figure is visualized in shades of grey, it is still readable (most people end up printing articles in bw to read them).
When it comes to integrating figures in the document, I mostly use Latex which outputs pdfs and deals quite well with figures in vector graphics. However the few times I had to use MSWord, I found such images difficult to integrate. I couldn't easily place a pdf or svg file in the text. I don't know if it is just me, but using the right tool for word processing with the right features is also quite important. Then journals often make a mess with manuscripts and figures, but unfortunately we cannot do much about it.
Note that figure width should also be adapted whenever possible, some journals accept to break the two-columns style for figures. This can be helpful for wide figures, and there is an option in Latex to specify that a figure should take the whole width of the page. I suggest using it when necessary and possible (but to use it sparingly still).
Finally my advice for those who work with programming/scripting languages to create figures (which you may cover in the next video), is to start by developing a template, and then to use it for every single figure in a project (article, report, thesis...). This helps immensely to obtain a consistent style. Every figure can then be tweaked as necessary, and it can be seen as some kind of safeguard against discrepancies. As a bonus, if you notice some styling error in all your graphs, it is likely that one slight modification in a single file will fix everything.
great video, very interesting, I am big into professional figures, very important, excited to see future videos
Awesome video, thank you for the effort, Andrey. Especially useful examples about decluttering!
vector grasphics are nicer than raster but if you have to do it right you will tikz it. This way you ensure correct font settings and get optimal alignment in your document. Latex has problems arranging svgs. A large amount of svg will cause performance issues, consider using pdfs in this case.
Wow, what a surprise to see a chilean among your collaborators
Great video !! Concise and helpful.
I’m really excited about this video and the rest of the series! I’m about to start my PhD, and I find this topic quite challenging, especially since there’s so little quality information available. Most resources only cover basic stuff like bar graphs or using Excel, but this tutorial is exactly what I needed! Thanks so much for all the effort you put into it :)
Appropriate. Stumbled upon this video. I am an Alum in Geoscience. Excellent
Great content bro. I am in Phd myself and found this very useful.
Thank u for your contribution. I'll be exciting to learn from this new video series.
Very nice explained and very helpful :) good work!
So I will generally agree with "no complimentary colors", however there are ways to muted the colors so that they don't clash. I'm a fan of a hazy blue and a hazy, slightly darker orange. blue and orange are the most colorblind friendly complimentary, they are the easiest on the eyes, and with some darkening/greying, they can look quite nice.
Wonderful talk, Kudos!!
في الوقت المناسب شكرا جزيلا على هذا الكورس الاكثر من رائع!
Really great info here! Thank you! Would love a follow up about how to practical implement some of these ideas (i.e. combined vector-raster figures) in python and latex.
This goes into my reference folder indefinitely.
This is what I need. Thank you!
A few days ago I had to give a presentation about a research project. Thanks to your video I now noticed some mistakes I made, especially the pixelated images. 😀👍🏻
Very good hints. Thanks.
This is very helpful. Thank you.
Can't fully agree with color advice...
Complementary colors are great when there are few strokes or spots which must be easy distinguishable. But it's important that colors don't become monochrome gray blob when viewed, converted or printed in gray-scale, so it's better to chose colors of different brightness.
I like the background ;)
Thanks Pouria 😂
Great video!
For colour palettes there are better tools that adjust for human perceptual colour difference. OKLAB pioneered it, if you want smooth gradients. LCH pickers are not bad. Some compensate for colour blindness too.
This is such valuable advice in general, should be mandatory in high school. You should review some of my published figures! Always open for public criticism haha
Hi Yury! I think it's a great idea to make videos on "let's improve figures provided by subscribers". I will announce this series when I have time 👍👍
Thank you for this great resource
looking forward! Awesome！
For color selection, there are a bunch of existing color blind palettes that are very legible and accessible !
Great video, great idea. The most overlooked aspect of science.
You make some very good points, but I think you lost it in the colors, a very hard and subjective matter, more often Red is not error, that is incorrect assumption. Although, red is usually the pen that a teacher use to correct and sometimes the color for the error bars, red more often means: fire, hot, much or the main point, it depends a lot on the subject and cultural experience. There is the classic blue to red for low to high values. In my opinion your plot should be black and transparent (gray) for the error figure. Also I believe nobody reads first the title of the figure, usually people read the plots like onions first the content and then the text that explains the content, layer by layer till the last detail. Please keep publishing more on the topic so we can all learn.
This is wonderful, I am writing a conference. I need ways to creating figures.
Hey Andrey, thank you for the most appreciated video, I have seen many videos in this regard but none of them was so detailed and pinpointed the actual problems I was confronted with. If I were to give a wish for your nexr video topic, to show us what software you use to create your unique graphs and diagrams? There are a lot of software like Excel, SPSS, Graphpad Prism, but what is your secret tip and how to get started, it would also fit alot to add the new A.I. tools that promise to solve your statistical problems and automatically generate charts and graphs.
Hi! Thanks for your feedback. Sure, I will show how I create figures.
My main tool is the Plots package in Julia. I also use CorelDRAW to adjust and modify the figures before putting them in the paper. Hopefully, I will manage to make the next tutorial in November-December.
Andrey
@chuscience thank you for the reply, I am looking forward to your next video.
Thank you sir!
I believe most students are familiar with this topic. Please introduce the best software for creating it, preferably one that is LaTeX-friendly.
I believe most students underestimate how much thought must be involved in creating good figures, but it could also be my own training which was lacking. There are entire books on the subjects!
Hello Andrey! Thank you so much, everything is amazingly explained. Can you tell please, is saving the illustrations in tiff format good choice? Or better to save in svg or pdf?
P.S I'm a physiologist and I use inkscape in making biological illustrations
Thanks Alex! I've never used tiff for my papers. As I see on Wikipedia, this is a raster graphics image format. This means that it will not save your results as scalable vector graphics. However, I don't know your images. Maybe tiff works well for your needs. Is tiff a popular format in your field?
Andrey
@@chuscience Yeah it is a popular format
@@chuscience Other thing I noticed when I saved my illustrations in PDF format. Even though I've drawn two lines which perfectly fits to each other, in PDF when you put the zoom on 75% one line appears longer than the other (they don't fit) but when you zoom in (100% and more) they fit to each other and appear in same length. It is so weird
After watching this video I only ask myself: how long do I need to go through every single video of this channel :0
Great video. How do you feel about ridgeline plots? I personally don't find them appealing, they always appear messy to me. Yet I see them everywhere. Second question, like almost 10% of all men I am red-green colourblind and so really struggle picking the right colours for my plots. Do you have any tips for me? Thanks!
Hi!
I haven't used Ridgeline plots in my papers, however I used Violin plots a few times. I think they are somewhat similar. Indeed, if we have too complex and messy data to present (say, many strange overlapping distributions), such plots are not the best choice. But they can look great in other cases. So my advice would be to test them and see if your figure looks nice with such distribution plots. Think about if readers will be able to understand the message of the figure.
I can't give great advice regarding picking colours for colourblind. Maybe select just one colour? Say, blue, and anything else is grey and black. Not the most intense blue though, maybe some shades of blue. So the main colour scheme would be a gradient from grey to blue, plus some black and white elements.
Andrey
@@chuscience Thanks! One more thing I noticed is that in density plots with pixelated grids my brain tells my eyes I have to focus, to make the picture sharper. That makes it unpleasant to look at for longer periods. For example, the plots shown in the AlgebraOfGraphics tutorial have this effect on me. I wonder if I'm the only one or if this is a known thing?
Something I had a discussion about with colleagues in the past: When (if ever) is it okay to trace over pixelated graphs with vector software, to give a nicer looking plot? Sometimes, you cannot (easily/realistically) get raw data out of a piece of measurement equipment for various reasons (I've even had this for security reasons before - you were not allowed to take measurements out, only the pngs of the plots).
In that case, I cannot get a nice vector graphic plot. Is it okay for me to trace the data in an vector graphics program? What about just drawing ledgends over it by hand, to make the graph more readable? Is it okay so long as I clearly state that it is a trace?
I've also done the same when adapting data from old figures, in papers that are only available as low-resolution scans. What about those cases?
Hi Vaes! I think that tracing pixelated data is fine as long as you do it in an ethical way. That is, you trace all data and don't manually modify it to hide some problems or to highlight only the things you want to show. If your ultimate goal is to make data nice and clear for readers, then tracing is a good idea. For this purpose, manually adding legends and extra notes should also be acceptable.
You can mention in a paper that the figure is traced and edited as a vector plot. For raw figures from the software, the readers can check your archive/repository [link].
Andrey
Could you consider in the next video on how did you import the figure into word... i think many students dont know how to import vector in the word file not only the issue of knowing them
Hi Salim! Thanks. I will try to cover it.
If you are willing to put effort for nice figures, I would highly recommend using Latex to prepare your paper. Word might be easy to use but (to me) has many limitations and tripping stones. Yes Latex has a learning curve and once figured out I can guarantee you will have better documents just from the text alone. In technical documents consitancy is important and automation aids this immensely which is (almost) entirely unavailable in Word (aside from the build in features). Already creating equations in Word is a mess.
Yes, I agree 👍 I switched completely to LaTeX for writing papers about 4 years ago.
thanks !
I am an Economics scholar but Engineering academics always have the best figures, to be honest.
Dear ChuScience , Can you recommend some professional fonts to use in academic publications
Hi! Great question. I don't think there is a single correct answer. You may need to test different fonts for your figures to see if they look nice in a manuscript. Some people advocate for Arial since it is simple and very well-known. Other options are Computer Modern (because of the LaTeX style), Georgia, Garamond, Courier, etc.
@@chuscience Thank you very much for your reply Mr.Andrey. This will be very helpful for me and many other academics.
Can you make a video what tools you use?
Hi! I will make tutorials about this. My main tools are Plots package for Julia Programming Language and CorelDRAW.
In the Windows system, PDF vector graphics cannot be directly inserted into Word and need to be converted to EMF format. The best way is to open it with Inkscape, making sure to use Cairo import instead of internal import, and then save it as .emf format. However, in the Mac OS system, PDF vector graphics can be directly inserted into Word with good results because Mac's Word converts PDFs into EMF format images. It is very strange that Microsoft's own Windows does not allow Word to smoothly insert PDF format vector graphics.
can you also share some of your fav books and research papers?
Thanks for the comment. I think that's a great idea to share/discuss some of my favourite books and papers. I will revise my literature list 👍
37:49 “Not enough red” lmao
36:10 those complimentary colors are WAY too similar, makes figures confusing.
I really hoped you would show the latex code for creating such figures 😔
Which is the name of your PDF viewer?
Hi! I was using a pretty standard Adobe Acrobat Reader, version 2024.003.20112. Sometimes I also open pdf in Google Chrome.
He sounds like Ilya Susketver
52:51 And then UML comes along and requires lile 1000 different arrows with completely different meanings.
this is so cute
As someone who does structural analysis, I never understood why programs use rainbow stress plot results for finite element analysis. It makes more sense to use 2-color gradients for positive and negative stress respectively. Doing this allows a sense of relative scale without having to reference the legend for every color.
raise your hand if you convert your png to 300x300 dpi
Read Tufte
Thanks! Will do.
Robinson Michelle Davis Brenda Martin Richard
The layout itself is outdated long time ago, should change to a webpage look with interactive code snippets already
It’s called a paper. Not a playground. If you want that you find in computer graphics or some areas of AI complementary websites that 10-20 people or so have worked on in addition to the manuscript. If you want that, put in the effort and contribute such playgrounds to your field of research.
Нет никакой теории цвета, кал это всё. Претенциозный и спекулятивный.
Fuck „research“ papers, I hate them
19:29 I bet if you make those numbers bigger then they will either obstruct the diagrams or they will be too bunched up together because there are a lot of numbers.
Yes, but this means the whole diagram type should be changed. Maybe it should be presented as a simplified graph with only branches and nodes, without generator symbols, substation buses, etc. Maybe no heatmap figure is needed here at all? What is the message here: that some locations have more capacity. Perhaps we should illustrate this distribution using some other plots. Or, if keeping the maps, they should be allocated much more space.