LMG's obsession with misleading graphs

[H0R53]

I've noticed during performance comparisons LMG uses cut axes, also known as truncated graphs. Here's a definition or two:

 

"In statistics, a misleading graph, also known as a distorted graph, is a graph that misrepresents data, constituting a misuse of statistics and with the result that an incorrect conclusion may be derived from it."

 

The context was most recently showcased during the Vega vs Pascal comparison. Several of the graphs had cut axes (data starting at a number other than zero).

 

"A truncated graph (also known as a torn graph) has a y axis that does not start at 0. These graphs can create the impression of important change where there is relatively little change."

 

Both of these graphs show the same data, yet are vastly different. This is what LMG is doing to misrepresent the difference between Vega and Pascal.

 

[MrMarriarty]

i love a good bar chart

[Enderman]

A lot of reviewers do this.

It's almost as if they weren't taught in school how to make proper graphs...

 

[H0R53]

Screens from the latest video showcase improperly scaled graphs. Notice how the scale is expanded to make differences far more outstanding than they actually are.

 

Apologies for the hyperwide screens, I don't feel like cropping all of them.

 

These particular graphs don't have truncated axes but quite a few in the older videos pre-"Petabyte NAS" have them.

[H0R53]

lmao those last 4 are super misleading, what's with the trailing data?

[Dredgy]

A truncated graph is perfectly acceptable to use, and makes visualising differences easier.

 

I actually saw a video on this recommended to me lately, and didn't realise anyone was against them, in fact most of my statistics graduates use them.

 

 

[GoodBytes]

I still prefer my charts. It follow more what the big (pretty much all) companies do:

 

But seriously, yes, I agree. My guess is that they did some of their charts too quickly. They probably quickly generated the charts in Excel and call it a day without much of a second look. It happens. With the rush to release new product review at the same time as everyone else, this is where sometimes you see stuff like this in reviews in general. But that is only my guess.

[H0R53]

1 minute ago, Dredgy said:

A truncated graph is perfectly acceptable to use, and makes visualising differences easier.

 

I actually saw a video on this recommended to me lately, and didn't realise anyone was against them, in fact most of my statistics graduates use them.

 

 

Well truncated graphs are bad, mmkay?

That video is trash, too.

 

2 minutes ago, GoodBytes said:

I still prefer my charts. It follow more what the big companies do:

 

  Reveal hidden contents

 

But seriously, yes, I agree. My guess is that they did some of their charts too quickly. They probably quickly generated the charts in Excel and call it a day without much of a second look. It happens. With the rush to release new product review at the same time as everyone else, this is where sometimes you see stuff like this in reviews in general.

Quality over quantity. Daily uploads are nice but at least make the data look more realistic

[The Belgian Waffle]

Or... Pause the video and look carefully at the graphs to fully understand it?

[H0R53]

Just now, The Belgian Waffle said:

Or... Pause the video and look carefully at the graphs to fully understand it?

That's not the point. Even if the video's paused, the graphs are still misleading.

[The Belgian Waffle]

Just now, H0R53 said:

That's not the point. Even if the video's paused, the graphs are still misleading.

So.. You are basically telling me that you cannot read a graph that doesn't start at (x;y)=(0;0)?

[Enderman]

1 minute ago, The Belgian Waffle said:

Or... Pause the video and look carefully at the graphs to fully understand it?

That's not the point of a graph.

A graph is supposed to give an accurate representation of the data without having to read the numbers.

 

AKA

if line A is twice as long as line B, that indicates that A is performing twice as well as B.

Not that A is 62fps and B is 61fps and the graph starts at 60.

That's called misleading data.

[H0R53]

Just now, The Belgian Waffle said:

So.. You are basically telling me that you cannot read a graph that doesn't start at (x;y)=(0;0)?

I can, but many, many people can't and don't understand graphs that aren't presented properly.

 

Stop trolling and go away.

[Dredgy]

Just now, H0R53 said:

Well truncated graphs are bad, mmkay?

That video is trash, too.

 

Well they're not, objectively. I have a lot,of graphs shoved in front of me on a daily basis and I'll always prefer something that gets the point across quickly. The right tool for the right job.

 

And that video, while simplistic, made the right points.

[Enderman]

1 minute ago, The Belgian Waffle said:

So.. You are basically telling me that you cannot read a graph that doesn't start at (x;y)=(0;0)?

Please learn what "misleading" means in terms of graphical representations of data.

https://en.wikipedia.org/wiki/Misleading_graph

http://www.statisticshowto.com/misleading-graphs/

 

Then maybe you will be able to make proper graphs.

[The Belgian Waffle]

Just now, Enderman said:

Please learn what "misleading" means in terms of graphical representations of data.

https://en.wikipedia.org/wiki/Misleading_graph

http://www.statisticshowto.com/misleading-graphs/

 

Then maybe you will be able to make proper graphs.

Thank you but I have a statistic class, and even if I am bad at statistics, I can read a graph

Just pause the damn video and open your damn eyes

[Kered124]

Not everyone is going to pause the video and take the time to realize that the bars aren't a linear representation of performance. 

I have to agree, I don't like the current representation of the data.

[H0R53]

2 minutes ago, Dredgy said:

Well they're not, objectively. I have a lot,of graphs shoved in front of me on a daily basis and I'll always prefer something that gets the point across quickly. The right tool for the right job.

 

And that video, while simplistic, made the right points.

 

1 minute ago, Enderman said:

Please learn what "misleading" means in terms of graphical representations of data.

https://en.wikipedia.org/wiki/Misleading_graph

http://www.statisticshowto.com/misleading-graphs/

 

Then maybe you will be able to make proper graphs.

I visited both of those sites before posting this thread, I googled "How to identify misleading graphs" and those were the top results.

[H0R53]

3 minutes ago, The Belgian Waffle said:

Thank you but I have a statistic class, and even if I am bad at statistics, I can read a graph

Just pause the damn video and open your damn eyes

 

3 minutes ago, Kered124 said:

Not everyone is going to pause the video and take the time to realize that the bars aren't a linear representation of performance. 

I have to agree, I don't like the current representation of the data.

^^^^^^^

 

Not everyone wants to pause the video.

 

The graphs are also poorly organized. When comparing two alike things, the bars representing data should be different colors. VEGA should have been yellow and pascal should have been blue, not 97th and average.

 

That's a really bad way to represent a comparison.

[GoodBytes]

For those who don't understand. my joke graph shows the problem of bad graphs.

 

You want the bars to shows properly the data, else use a table.

The point of a graph is to make it easy to understand over than a just a table with numbers, as mentioned by others.

 

On my joke graph I highlight the problem with 1 fps performance difference between 2x CPUs. You would expect that the 2 lines are close to each other, but here you see them far apart. And you don't want the reverse. Where, the value should be far apart, but you hide it by having your measurement steps so large that it looks like the bars are identical level.

 

This kind of bad graph design could indicate bias, even though it might not be intentional. Meaning, say for example, you compare 2 GPUs, one from company A and the other from Company B. You are a company A fanboy, and the GPU performance of A is under B's. So you trick the measurement to show that there is little performance difference in your graph, and act like it is minor, even though it is large. And on the opposite side you can do like my joke graph where you stretch the graph to insane amount to show 1 fps, acting like it is huge, and you should get company B product, even taught A cost less and is "1 fps less".

 

While Excel auto-generate graphs, where it usually does a decent job (assuming you know how to fill up a table in the first place), it is an automated process at the end of the day. It is up to you to fix your axis after the generated graph if you see a problem.

 

The reader should not have to pause the video and do an analysis on the graph, else just give the table data.

[Enderman]

1 hour ago, The Belgian Waffle said:

Thank you but I have a statistic class, and even if I am bad at statistics, I can read a graph

Just pause the damn video and open your damn eyes

Again, that's not the point of having a graph.

At that point you may as well just list the numbers in a spreadsheet and not put any graphics on the screen.

 

There are people who do not take statistics and may not be aware of misleading graphs or don't read the numbers on every single graph if they just want to get a quick idea of the performance.

A misleading graph will give people who skim over the video the wrong idea.

[LogicalDrm]

Graphs are just part of why I don't really "watch" LTT reviews on CPUs and GPUs. I listen them as preview and watch OC3D or Jayz for real review. OC3D having benefit of written review to look for graphs. They might not be perfect either. But they usually have many generations of products on the graph so cheating isn't so tempting.

[Blind Krill 796]

3 hours ago, GoodBytes said:

The reader should not have to pause the video and do an analysis on the graph, else just give the table data.

LOL at this.  I have to pause the video every time a graph comes up since they are only displayed for 2 seconds.... 

[PlayStation 2]

I see both using truncated graphs and using more accurate graphs that cover most everything in the range. If you only use truncated graphs to make a difference seem much larger than it actually is, then that would be 100% misleading. However, if you used a combination of both, I'd be perfectly fine.

[themctipers]

4 hours ago, H0R53 said:

lmao those last 4 are super misleading, what's with the trailing data?

i dont like the last 4 graph categories. they seem misleading, and hard to understand for me, and is useless. the review should tell me if the CPU/GPU is good value against its competitors and pricing, don't give me a ??? what the fuck graph that might tell me that if this gpu was $200 it will be slightly better than if a 1070 was $200