I've lost a bit more than a hundred (yes, 100) pounds during the last year. My social skills have improved due to an increase in my confidence.
But here is the interesting part.
People do treat me much better. I get more respect. More smiles. They think I am more intelligent, and show this by listening to what I have to say.
And even though I am an introvert, and somewhat shy, people tell me how "social" I am.
I have not changed the way I dress (jeans, t-shirt, sneakers). Though when I'm wearing exercise clothing (running shorts, a sleeveless shirt, running shoes), people are more likely to approach me.
So, this person does have a good point. We are indeed judged by our appearances. In fact, our social standing is calculated on the fly by others on how we look.
Last time something from judg.me was posted, there was a lot of criticism about your statistical methods of analysis. You've clearly put a lot of effort into attempting to do a better analysis of the data, and this time there's certainly an improvement in terms of actual analytical methods used.
However, the graphs (as others have said) are terrible. Simply having a graph type available to you doesn't mean you have to use it! For the sake of your readers, please, please, only use one type of graph. Stick with a plain bar chart. Nobody minds that it's not as fancy looking as a pseudo-3D bar chart or one with pointed cones. People want to see the data, not the fancy chart.
An additional point to this is labelling your axes. Please do it! If I can't see clearly what an axis represents, I'm inclined to ignore your data. Regardless of how interesting your data is, I'm going to stop reading if it's presented terribly.
These are the sorts of things that get hammered into you in scientific subjects at school/college. You don't need to do that type of class to know them though. There's plenty of information online about this stuff.[1] Worth noting is section 2.6, "Do not misuse the tools". Bad graphs are to statistics what the <blink> tag is to HTML.
Please consider showing more of the variation and including more information about it in your results.
You conclude, for example, that "In general, it can be concluded that people can and do make snap judgments about personality and abilities based on superficial physical characteristics."
However, a cursory stroll through ranking people on your site reveals that actually, peoples' "snap judgments" show very high subjective variability, and you see these graphs which just have data points all over the place, looking more like a uniform distribution than a clear "everyone things you look smart and extroverted."
Given this, can you profile a burst of ratings? Can you say that there are a coherent group who rated Alice as smart extrovert and Bob as smart introvert and Carol as dumb extrovert -- and can you profile their preferences? Or do most people just browse through your site clicking randomly on the axes without any real reflection on the people they're rating?
We could see this if we could see error bars, but aside from your very first image where you try to show that there is a negative correlation between smartness and sociability, info on variability seems quite nonexistent. (And even there, it is clear that the variability dwarfs the trendline.) Also, you might want to see if there are systematic biases -- what happens if you flip the introvert/extrovert axis for some IP addresses?
You claim to have proven that people make snap judgments based on superficial physical appearance. But it's based on a test in which you explicitly demanded that the test subjects do so. Failure to do so meant non-participation in your test. How can you claim to have tested the hypothesis?
When showing similar types of data try to use the same type of graph. Once the reader has learnt to interpret results from one graph they can easily apply it to the next. It looks like you've cycle through every graph type Excel has to offer.
Used to have a terrible posture until it gave me back problems. Spent a week correcting it only to be shocked at the difference it made to my attractiveness to women - wish I'd known earlier '-(
The research is a good idea, an investigation of the halo effect. The population of graders wasn't mentioned, that can have a huge effect on the result. Also, was 1000 people enough to analyze 12 variables?
However, I have to say, the graphs are ugly. Eg, spikes aren't good, you could have gone with with a bar chart. Also, 3d graphs are unneeded, and in general are very hard to do right. They can't convey much information. With just 2x2 variables, there are 4 datapoints. That would be fine on a 2d picture.
[+] [-] rmATinnovafy|14 years ago|reply
I've lost a bit more than a hundred (yes, 100) pounds during the last year. My social skills have improved due to an increase in my confidence.
But here is the interesting part.
People do treat me much better. I get more respect. More smiles. They think I am more intelligent, and show this by listening to what I have to say.
And even though I am an introvert, and somewhat shy, people tell me how "social" I am.
I have not changed the way I dress (jeans, t-shirt, sneakers). Though when I'm wearing exercise clothing (running shorts, a sleeveless shirt, running shoes), people are more likely to approach me.
So, this person does have a good point. We are indeed judged by our appearances. In fact, our social standing is calculated on the fly by others on how we look.
Edited to fix spelling disaster.
[+] [-] metabrew|14 years ago|reply
I kept wondering how anyone who chooses hard-to-read shonky 3d bar charts like that could possibly perform a decent statistical analysis.
[+] [-] wr1472|14 years ago|reply
Unless of course you judge the validity of a study based on the appearance of its graphs?
[+] [-] bvi|14 years ago|reply
[+] [-] leoedin|14 years ago|reply
However, the graphs (as others have said) are terrible. Simply having a graph type available to you doesn't mean you have to use it! For the sake of your readers, please, please, only use one type of graph. Stick with a plain bar chart. Nobody minds that it's not as fancy looking as a pseudo-3D bar chart or one with pointed cones. People want to see the data, not the fancy chart.
An additional point to this is labelling your axes. Please do it! If I can't see clearly what an axis represents, I'm inclined to ignore your data. Regardless of how interesting your data is, I'm going to stop reading if it's presented terribly.
These are the sorts of things that get hammered into you in scientific subjects at school/college. You don't need to do that type of class to know them though. There's plenty of information online about this stuff.[1] Worth noting is section 2.6, "Do not misuse the tools". Bad graphs are to statistics what the <blink> tag is to HTML.
[1, The UNs guide to presenting statistics]: http://www.unece.org/fileadmin/DAM/stats/documents/writing/M...
[+] [-] bvi|14 years ago|reply
[+] [-] drostie|14 years ago|reply
You conclude, for example, that "In general, it can be concluded that people can and do make snap judgments about personality and abilities based on superficial physical characteristics."
However, a cursory stroll through ranking people on your site reveals that actually, peoples' "snap judgments" show very high subjective variability, and you see these graphs which just have data points all over the place, looking more like a uniform distribution than a clear "everyone things you look smart and extroverted."
Given this, can you profile a burst of ratings? Can you say that there are a coherent group who rated Alice as smart extrovert and Bob as smart introvert and Carol as dumb extrovert -- and can you profile their preferences? Or do most people just browse through your site clicking randomly on the axes without any real reflection on the people they're rating?
We could see this if we could see error bars, but aside from your very first image where you try to show that there is a negative correlation between smartness and sociability, info on variability seems quite nonexistent. (And even there, it is clear that the variability dwarfs the trendline.) Also, you might want to see if there are systematic biases -- what happens if you flip the introvert/extrovert axis for some IP addresses?
[+] [-] msellout|14 years ago|reply
[+] [-] MattArnold|14 years ago|reply
[+] [-] rickyconnolly|14 years ago|reply
[+] [-] shutton|14 years ago|reply
Other than that, quite interesting!
[+] [-] Mordor|14 years ago|reply
[+] [-] sk5t|14 years ago|reply
[+] [-] rockmeamedee|14 years ago|reply
However, I have to say, the graphs are ugly. Eg, spikes aren't good, you could have gone with with a bar chart. Also, 3d graphs are unneeded, and in general are very hard to do right. They can't convey much information. With just 2x2 variables, there are 4 datapoints. That would be fine on a 2d picture.
Otherwise keep on with the research...
[+] [-] Paul_S|14 years ago|reply
Also, asking people a direct question will often get you a dishonest answer. Maybe there's a way to ask the same question in an indirect way?
[+] [-] empthought|14 years ago|reply
[+] [-] frankiewarren|14 years ago|reply
[+] [-] bvi|14 years ago|reply
[+] [-] schme|14 years ago|reply
[+] [-] jcmhn|14 years ago|reply
[+] [-] doubleconfess|14 years ago|reply