1945:"Scientific Paper Graph Quality"

This forum is for the individual discussion thread that goes with each new comic.

Moderators: Moderators General, Prelates, Magistrates

User avatar
cellocgw
Posts: 1898
Joined: Sat Jun 21, 2008 7:40 pm UTC

1945:"Scientific Paper Graph Quality"

Postby cellocgw » Mon Jan 22, 2018 4:07 pm UTC

Image

Title-text: "The worst are graphs with qualitative, vaguely-titled axes and very little actual data"

I propose there should be a brief up-spike in "goodness" around 1985 as millions of techies switched to MacDraw to present figures and charts.
https://app.box.com/witthoftresume
Former OTTer
Vote cellocgw for President 2020. #ScienceintheWhiteHouse http://cellocgw.wordpress.com
"The Planck length is 3.81779e-33 picas." -- keithl
" Earth weighs almost exactly π milliJupiters" -- what-if #146, note 7

User avatar
JGeezer
Posts: 54
Joined: Wed May 01, 2013 6:00 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby JGeezer » Mon Jan 22, 2018 4:08 pm UTC

This is definitely one of GLR's self-referential efforts.
Image All you molpys, get off my lawn!

User avatar
Izawwlgood
WINNING
Posts: 18686
Joined: Mon Nov 19, 2007 3:55 pm UTC
Location: There may be lovelier lovelies...

Re: 1945:"Scientific Paper Graph Quality"

Postby Izawwlgood » Mon Jan 22, 2018 4:09 pm UTC

"Methods: Graphs were drawn with MSPaint and PhotoShop"

Noooooooooooo
... with gigantic melancholies and gigantic mirth, to tread the jeweled thrones of the Earth under his sandalled feet.

User avatar
Reka
Posts: 198
Joined: Thu Sep 20, 2012 10:21 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby Reka » Mon Jan 22, 2018 4:24 pm UTC

I don't get where MSPaint enters into this: surely anyone who's using Powerpoint would have Excel available to them, and would use Excel to generate the charts?

bottles
Posts: 3
Joined: Wed Dec 01, 2010 7:05 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby bottles » Mon Jan 22, 2018 4:29 pm UTC

Reka wrote:I don't get where MSPaint enters into this: surely anyone who's using Powerpoint would have Excel available to them, and would use Excel to generate the charts?


That's a really lovely worldview.

DanD
Posts: 302
Joined: Tue Oct 05, 2010 12:42 am UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby DanD » Mon Jan 22, 2018 4:35 pm UTC

Reka wrote:I don't get where MSPaint enters into this: surely anyone who's using Powerpoint would have Excel available to them, and would use Excel to generate the charts?


You're assuming A)competence. And B) actual data.

User avatar
cellocgw
Posts: 1898
Joined: Sat Jun 21, 2008 7:40 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby cellocgw » Mon Jan 22, 2018 4:41 pm UTC

Reka wrote:I don't get where MSPaint enters into this: surely anyone who's using Powerpoint would have Excel available to them, and would use Excel to generate the charts?


I have seen, in documents presented to customers, charts generated with Excel which showed curves going backwards. Yep, some smart guy (i.e. not) saw the 'smoother' tools in Excel and ran some ultra-splined option on noisy data points.
https://app.box.com/witthoftresume
Former OTTer
Vote cellocgw for President 2020. #ScienceintheWhiteHouse http://cellocgw.wordpress.com
"The Planck length is 3.81779e-33 picas." -- keithl
" Earth weighs almost exactly π milliJupiters" -- what-if #146, note 7

User avatar
balthasar_s
Posts: 2236
Joined: Mon Nov 28, 2011 3:20 pm UTC
Location: secret base on the Moon
Contact:

Re: 1945:"Scientific Paper Graph Quality"

Postby balthasar_s » Mon Jan 22, 2018 6:12 pm UTC

"Scientific Paper Graph Quality"?

I'm more interested in the scientific graph paper quality.
BFTF replay
Good luck, my blitzing friends!
BTTBAA:1023 # Mustard? Use the mirror! Blitzing? Also use the mirror! And here's why. # OTT facebug copy
that's a robot so it doesn't countImage
This text was autogenerated:
swimming unit detector active

User avatar
orthogon
Posts: 2904
Joined: Thu May 17, 2012 7:52 am UTC
Location: The Airy 1830 ellipsoid

Re: 1945:"Scientific Paper Graph Quality"

Postby orthogon » Mon Jan 22, 2018 6:30 pm UTC

balthasar_s wrote:"Scientific Paper Graph Quality"?

I'm more interested in the scientific graph paper quality.

OMY! Graph paper! That totally used to be a thing!
xtifr wrote:... and orthogon merely sounds undecided.

User avatar
Soupspoon
You have done something you shouldn't. Or are about to.
Posts: 3391
Joined: Thu Jan 28, 2016 7:00 pm UTC
Location: 53-1

Re: 1945:"Scientific Paper Graph Quality"

Postby Soupspoon » Mon Jan 22, 2018 6:36 pm UTC

I still have some Log-Log graph paper. I'm telling you this so that it's on record that there exists some dead tree still awaiting use, except for some corporate imagery doodlings. So please log my ex-log log-log logo, before you log off for the evening.

User avatar
yakkoTDI
Posts: 47
Joined: Fri Jul 10, 2015 4:02 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby yakkoTDI » Mon Jan 22, 2018 10:42 pm UTC

orthogon wrote:
balthasar_s wrote:"Scientific Paper Graph Quality"?

I'm more interested in the scientific graph paper quality.

OMY! Graph paper! That totally used to be a thing!


I love the idea of graph paper. Sadly my handwriting looks like crap unless it is tiny and those graph boxes are just too big. :(

airdrik
Posts: 236
Joined: Wed May 09, 2012 3:08 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby airdrik » Mon Jan 22, 2018 11:28 pm UTC

yakkoTDI wrote: <snipped quotes>
I love the idea of graph paper. Sadly my handwriting looks like crap unless it is tiny and those graph boxes are just too big. :(

How big is too big?
You can (could) get (print) graph paper in a wide variety of sizes. Most "standard" graph paper you get has sizes akin to that of standard lined paper, but there are also varieties with graph sizes down to 1/10" and 1 mm, and if you have the ink available you can print whatever size you want (shouldn't require much since you'll want to print lines that are visible enough but largely get out of the way of what you are graphing/drawing/writing).
I recall finding some with a small grid (probably 1/10") that I used to doodle fractals on back in my school days.

User avatar
somitomi
Posts: 601
Joined: Fri Nov 06, 2015 11:21 pm UTC
Location: can be found in Hungary
Contact:

Re: 1945:"Scientific Paper Graph Quality"

Postby somitomi » Tue Jan 23, 2018 12:11 pm UTC

orthogon wrote:
balthasar_s wrote:"Scientific Paper Graph Quality"?

I'm more interested in the scientific graph paper quality.

OMY! Graph paper! That totally used to be a thing!

What do you mean "used to"? I actually have some right here in my desk, because I needed them regularly a couple semesters ago.
—◯-◯

pierreb
Posts: 15
Joined: Fri Apr 29, 2011 8:20 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby pierreb » Tue Jan 23, 2018 2:23 pm UTC

I can't help but notice the decrease in quality started BEFORE the powerpoint/ms-paint era.

It seems clear to me the quality decline wasn't MSFT made.

I'm a "graph quality change" denier.

Yu_p
Posts: 51
Joined: Wed Oct 12, 2011 9:00 am UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby Yu_p » Tue Jan 23, 2018 5:32 pm UTC

pierreb wrote:I can't help but notice the decrease in quality started BEFORE the powerpoint/ms-paint era.
It seems clear to me the quality decline wasn't MSFT made.
I'm a "graph quality change" denier.


There is a very real chance, that the decline was started simply by using software instead of drawing graphs by hand, as lowering the entry barrier to something typically results in more people doing it, with on average less knowledge about it. Supposedly that effect (with regard to GUI design) was part of what damaged the reputation of Tcl/Tk.

Wasn't it even common to have a dedicated "Grapher" as an employee for publication quality graphs? Kind of like "computer" used to be a job description. I fear the day when people think of a machine, when they hear "scientist" :shock: Or should I be more worried that by that time the same applies to the term "people"?

User avatar
GlassHouses
Posts: 116
Joined: Thu Nov 24, 2016 12:41 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby GlassHouses » Tue Jan 23, 2018 7:32 pm UTC

Yu_p wrote:
pierreb wrote:I can't help but notice the decrease in quality started BEFORE the powerpoint/ms-paint era.


There is a very real chance, that the decline was started simply by using software instead of drawing graphs by hand, as lowering the entry barrier to something typically results in more people doing it, with on average less knowledge about it.


The first time I wrote a report on a computer, in college in 1985 or '86, I struggled with my graphs. I found it easy to draw good graphs by hand, with graphing paper, a ruler, and a calculator, but on the computer I was lost. It wasn't until years later, having learned how to do graphs in Mathematica and, later, Excel, that the combination of "computer" and "graph" slowly stopped giving me anxiety and nausea.

In other words, my explanation for the phenomenon (which I'm not convinced exists, but whatever) isn't that graphing became too easy, but too hard. I still can't use gnuplot without a bottle of Valium.

Yu_p
Posts: 51
Joined: Wed Oct 12, 2011 9:00 am UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby Yu_p » Tue Jan 23, 2018 8:10 pm UTC

GlassHouses wrote:In other words, my explanation for the phenomenon (which I'm not convinced exists, but whatever) isn't that graphing became too easy, but too hard. I still can't use gnuplot without a bottle of Valium.


Likely a matter of education likely. I haven't really learned doing proper graphs by hand to begin with, so I would be lost when trying to do stuff like drawing a smooth curve based on actual data. I understand that there are some tools used to pull it off, but I don't think I've ever even seen them.

Sketches of course, are much easier by hand. But production quality graphs? No idea how to do them that way. Funny enough, I still find gnuplot the most intuitive solution :) I even wrote a simple (unpublished) wrapper for python, because I had grown too annoyed by alternatives.

Last time I tried Excel it crashed from too-many-datapoints though :)
Last edited by Yu_p on Wed Jan 24, 2018 10:48 am UTC, edited 1 time in total.

xtifr
Posts: 322
Joined: Wed Oct 01, 2008 6:38 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby xtifr » Wed Jan 24, 2018 12:44 am UTC

One thing I noticed is that at just about the same time that TeX came along and made the fonts and layouts of papers look much nicer, gnuplot came along and made the fonts and labels on graphs look much worse.
"[T]he author has followed the usual practice of contemporary books on graph theory, namely to use words that are similar but not identical to the terms used in other books on graph theory."
-- Donald Knuth, The Art of Computer Programming, Vol I, 3rd ed.

User avatar
svenman
Posts: 564
Joined: Fri Jun 14, 2013 2:09 pm UTC
Location: 680 km NNE of the Château d'If

Re: 1945:"Scientific Paper Graph Quality"

Postby svenman » Thu Jan 25, 2018 1:10 pm UTC

orthogon wrote:
balthasar_s wrote:"Scientific Paper Graph Quality"?

I'm more interested in the scientific graph paper quality.

OMY! Graph paper! That totally used to be a thing!

By an odd coincidence (I assume), graph paper is also mentioned in the latest SMBC strip.

Either that, or Zach Weinersmith secretly lurks on the xkcd forums and draws the occasional inspiration from here.
Mostly active on the One True Thread.
If you need help understanding what's going on there, the xkcd Time Wiki may be useful.

Addams didn't die! But will Addams have a place to live? You can help!

Randallspeed to all blitzers on the One True Thread!

bottles
Posts: 3
Joined: Wed Dec 01, 2010 7:05 pm UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby bottles » Fri Jan 26, 2018 4:42 am UTC

Yu_p wrote:
GlassHouses wrote:In other words, my explanation for the phenomenon (which I'm not convinced exists, but whatever) isn't that graphing became too easy, but too hard. I still can't use gnuplot without a bottle of Valium.


Likely a matter of education likely. I haven't really learned doing proper graphs by hand to begin with, so I would be lost when trying to do stuff like drawing a smooth curve based on actual data. I understand that there are some tools used to pull it off, but I don't think I've ever even seen them.

Sketches of course, are much easier by hand. But production quality graphs? No idea how to do them that way. Funny enough, I still find gnuplot the most intuitive solution :) I even wrote a simple (unpublished) wrapper for python, because I had grown too annoyed by alternatives.

Last time I tried Excel it crashed from too-many-datapoints though :)


The trick to hand-drawn graphs is that you linearize the data. Rather than plotting pH vs [HA] and trying to figure out where the middle of the S curve is (to get the pKa), you plot pH vs log([HA]/[A-]) and use a straightedge to get the line. Then finding the pKa is just a matter of seeing where the line intersects an axis. There's a lot of situations where we don't bother doing the linearization part anymore because modern graphing software (Excel if you're a wizard, but more like graphpad or kaleidagraph) has no problem fitting that sort of curve.

User avatar
orthogon
Posts: 2904
Joined: Thu May 17, 2012 7:52 am UTC
Location: The Airy 1830 ellipsoid

Re: 1945:"Scientific Paper Graph Quality"

Postby orthogon » Fri Jan 26, 2018 11:04 am UTC

bottles wrote:The trick to hand-drawn graphs is that you linearize the data. Rather than plotting pH vs [HA] and trying to figure out where the middle of the S curve is (to get the pKa), you plot pH vs log([HA]/[A-]) and use a straightedge to get the line. Then finding the pKa is just a matter of seeing where the line intersects an axis. There's a lot of situations where we don't bother doing the linearization part anymore because modern graphing software (Excel if you're a wizard, but more like graphpad or kaleidagraph) has no problem fitting that sort of curve.


There are (at least) three different reasons for drawing a graph:

1. To enable some quantity, e.g. a constant of proportionality, to be extracted from noisy data using the human eye-brain combination
2. To see how well the data supports the hypothesised relationship
3. To illustrate visually the relationship between variables

The linearisation technique is good for (1) and (2), since it's easier for a human to eyeball the best straight-line fit, and it's easy to see how well the data points cluster around that line. Straight lines are something we've evolved to see, they can be generated easily with a ruler, and the same ruler will work whatever the parameters (gradient and intercept). Once you get into curves, there are too many parameters and you start to need French Curves (remember them!) and the like. As you say, the best fit (1) can be found in an objective way, as can the quality of the fit (2); it's now very easy for computers to do this, and given how open to abuse data can be, you really want to be using an objective method. The graph can still be useful as a visual way to illustrate the quality of fit.

Linearisation is essentially the opposite of what you want for (3), since the whole point there is to see whether the curve flattens off, asymptotes, peaks, where it's steepest, etc. Showing that some function of the variables results in a straight line literally removes all the interesting features from the relationship; it tells you no more than the hypothesised mathematical equation.
xtifr wrote:... and orthogon merely sounds undecided.

Yu_p
Posts: 51
Joined: Wed Oct 12, 2011 9:00 am UTC

Re: 1945:"Scientific Paper Graph Quality"

Postby Yu_p » Fri Jan 26, 2018 1:37 pm UTC

orthogon wrote:Linearisation is essentially the opposite of what you want for (3), since the whole point there is to see whether the curve flattens off, asymptotes, peaks, where it's steepest, etc. Showing that some function of the variables results in a straight line literally removes all the interesting features from the relationship; it tells you no more than the hypothesised mathematical equation.


Most of the graphs I encounter a spectra of some kind. For those you are interested in where they have peaks, how wide and high the peaks are, and sometimes, what shapes the peaks have. And whether what looks like a single peak, actually is a single peak or multiple ones close together. This is normally done by fitting a sum of modeled peaks to the data, and then looking at the (possibly smoothed) plot of the raw data to check if the fit has found a proper global minimum, and if it explains all the features of the data.

Honestly, I wasn't even thinking of using graphs as the tool the extract data from a graph (except for qualitative understanding). Cases were I was using that came up sometimes in private interest stuff though, mostly when data is being grossly misinterpted for selling stuff (e.g. that claim that SSDs will eventually be cheapter than HDDs (doesn't match the data), or that tablets/PCs were supposed to have vanished by now).

Image


Return to “Individual XKCD Comic Threads”

Who is online

Users browsing this forum: No registered users and 37 guests