mma12注册机_mma战斗机的手臂和腿长之间的相关性

mma12注册机

I recently published a paper on armspan and fighting success in mixed martial arts (MMA) fighters. I found that fighters with greater armspans have a very small advantage in fighting i.e. they win more. One issue I ran into was that I couldn’t rule out the hypothesis that leg length was driving the effect, because leg length data is not routinely collected for MMA fighters. We’d expect fighters with longer arms to have longer legs too, which would also presumably provide an advantage. I couldn’t find any data on correlations between arm and leg length in fighters, so I just left it at that.

我最近发表了一篇关于混合武术(MMA)战斗机的臂展和成功战斗的论文 。 我发现，臂展更大的战斗机在战斗中的优势很小，即获胜更多。 我遇到的一个问题是，我不能排除腿长导致这种效果的假说，因为腿长数据不是常规为MMA战斗机收集的。 我们希望双臂较长的战斗机也能拥有更长的腿，这大概也会提供优势。 我找不到关于战斗机手臂和腿长之间的相关性的任何数据，所以我就把它留在那儿了。

I recently found an MMA league called The professional fighters league. It has armspan and leg length data from about 78 fighters. I thought it might be interesting to look at the correlation between arm and leg length in MMA fighters. It’s not a large sample, so doing anything fancier than a correlation is unlikely to be informative (my paper had ~1600 fighters!). We’d expect a large correlation between arm and leg length, so 78 fighters is very likely good enough. For example, a simple power analysis using free a free program Gpower tells us that to have a 99% chance of detecting a correlation of |0.6| or larger (if it exists at all) only requires 35 data points, 30 if you have a 1-tailed hypothesis as we do. 78 fighters with a 1-tailed test should give us 99% power to detect correlations as low as |0.41|.

我最近发现了一个MMA联赛，叫做职业战士联赛 。 它具有来自约78名战士的臂展和腿长数据。 我认为研究MMA战斗机的臂长与腿长之间的相关性可能很有趣。 这不是一个大样本，所以做任何比相关性更出色的事情都不会提供信息(我的论文有大约1600名战士！)。 我们希望手臂和腿的长度之间存在很大的相关性，因此78名战斗机很可能足够好。 例如，使用免费的免费程序Gpower进行的简单功率分析告诉我们，有99％的机会检测到| 0.6 |的相关性。 或更大(如果有的话)仅需要35个数据点，如果您像我们一样有1尾假设，则只需30个数据点。 78架经过1尾测试的战斗机应该给我们99％的力量来检测低至| 0.41 |的相关性。

I manually went through all fighter pages and wrote down the armspan and leg length data into Excel. The data can be found on the open science foundation page for my armspan article (I didn’t really know where else to put it). I did a simple analysis and thought heck, why not make this into an R tutorial for the interested!

我手动浏览了所有战斗机页面，并将臂展和腿长数据记入Excel。 可以在我的臂膀文章的开放科学基金会页面上找到这些数据 (我真的不知道该把它放在哪里)。 我做了一个简单的分析，并想了一下，为什么不把它变成有兴趣的R教程！

First we download the data from my OSF directory, and read it into R:

首先，我们从OSF目录下载数据，并将其读入R：

download.file(‘https://osf.io/4d2pv/download','arm_leg_PFL.csv')
data= read.csv(‘arm_leg_PFL.csv’)

First let’s make histograms for the 2 dimensions, to see how they’re distributed, how they vary (remember you’ll need these packages installed first!). We use the GGplot2 package as it gives nice looking graphs, and the gridExtra package as it allows us to present multiple graphs simultaneously.

首先，让我们为这两个维度制作直方图，以了解它们的分布方式以及它们的变化方式(请记住，您首先需要安装这些软件包！)。 我们使用GGplot2程序包，因为它可以提供漂亮的图形，而使用gridExtra程序包，因为它可以使我们同时显示多个图形。

library(ggplot2)
library(gridExtra)plot1 = ggplot(data, aes(x = armspan))+
geom_histogram(binwidth = 1, fill = ‘darkolivegreen’)+
ggtitle(‘Histogram of fighter armspans (inches)’)plot2 = ggplot(data, aes(x = leg_length)) +
geom_histogram(binwidth = 1, fill = ’navy’)+
ggtitle(‘Histogram of fighter leg lengths (inches)’)grid.arrange(plot1, plot2, nrow=1)

The variables are a bit skewed, not as normally distributed as we’d like, or expect in a general population sample (body proportions tend to be normally distributed). Many statistical methods work best when your data is normally distributed. That said, neither variable seems to have any extreme values that could pose problems for analysis, though some beast has 47-inch legs!

这些变量有些偏斜，不像我们期望的那样呈正态分布，也不像一般人群样本中所期望的那样(身体比例倾向于呈正态分布)。 当数据呈正态分布时，许多统计方法效果最佳。 就是说，尽管有些野兽有47英寸的腿，但两个变量似乎都没有任何极高的值可能引起分析问题！

The very different scales of the 2 measurements is probably because armspan is measured including the length of both arms and chest width (the distance between middle fingertips with arms outstretched), whereas leg length here is probably something like distance from hip bone to foot for 1 leg. The ratings seem a bit high for inseam measurements, as most have measurements of 36+, which would be a decent leg length for a 6ft (1.83 m) man, and many of these fighters are much shorter than that. But who knows.

两次测量的比例非常不同，可能是因为测量了臂展，包括双臂的长度和胸部的宽度(双臂伸出的中指之间的距离)，而此处的腿长可能类似于髋骨到脚的距离1腿。 在下in骨测量中，该等级似乎有点高，因为大多数人的测量值为36+，这对于6英尺(1.83 m)的男人来说是一个体面的腿长，其中许多战士比这短得多。 但是谁知道。

Now we plot them both against each other. Geom_point adds the data points and geom_smooth adds a trend line. method = ‘lm’ indicates we want a straight line, but there are other options.

现在，我们将它们相互绘制。 Geom_point添加数据点，geom_smooth添加趋势线。 method ='lm'表示我们想要一条直线，但是还有其他选择。

ggplot(data, aes(x = leg_length, y = armspan))+
geom_point(colour = ‘slateblue’)+
geom_smooth(method=’lm’)+
ggtitle(‘Scatter plot of fighter leg length vs armspan’)+
theme(plot.title = element_text(hjust = 0.5))+
xlab(‘Leg length (inches)’)+
ylab(‘Armspan (inches)’)

The ‘theme’ bit of code just centres the title (aesthetic preference)

代码的“主题”位只是将标题居中(审美偏好)

Images by author

图片作者

They’re looking pretty well correlated. But there more to science than graphs! Let’s test it more formally. We use both a Pearson and a Spearman's correlations, as the former has more power, the latter is more robust to outliers. If they differ wildly, it might indicate that outliers are messing with our analyses (though judging by the graph, this is unlikely).

他们看起来相关性很好。 但是科学不仅仅是图！ 让我们对其进行更正式的测试。 我们同时使用Pearson和Spearman的相关性，因为前者具有更大的功效，而后者则对异常值具有更强的鲁棒性。 如果它们之间存在巨大差异，则可能表明离群值正在干扰我们的分析(尽管通过图表判断，这不太可能)。

cor.test(data$armspan, data$leg_length, method = ’pearson’, alternative = 'greater')
cor.test(data$armspan, data$leg_length, method = ’spearman’, alternative = 'greater')

Note that if you leave out the ‘method’ argument R will just do a Pearson one. ‘Alternative = greater’ means that we have a 1 tailed hypothesis that they are positively correlated.

请注意，如果省略“方法”参数，R只会做一个Pearson。 “替代=更大”意味着我们有一个1尾的假设，认为它们是正相关的。

The correlations don’t differ much, so outliers are not really messing things up. The Pearson correlation is +0.47, and the Spearman +0.48. Both have P values < 0.001, highly significant. Given that our sample size is certainly large enough to detect these correlations, the extremely low p-values aren’t surprising.

相关性相差不大，因此离群值并没有真正弄乱事情。 皮尔逊相关系数为+0.47，而斯皮尔曼系数为+0.48。 两者的P值均<0.001，非常显着。 鉴于我们的样本量肯定足够大，可以检测到这些相关性，因此p值极低并不奇怪。

Interestingly r = +0.47 is a bit lower than I’d have predicted! For context, The correlation between armspan and height is around r = 0.8. There could be a few reasons for this, such as measurement error, especially if fighters or their teams did the measuring and reported it to the league. This will introduce noise which reduces the magnitude of any correlations. Armspan also includes chest width, whereas the leg measurements seem to reflect the measurement of a single leg. If we had arm length, that might showed a larger correlation with leg length. But alas, in the real world you rarely have perfect data, especially when you scrape it from the internet.

有趣的是，r = +0.47低于我的预期！ 就上下文而言，臂展与身高之间的相关性约为r = 0.8。 可能有一些原因，例如测量错误，尤其是如果战士或他们的团队进行了测量并将其报告给联盟。 这将引入噪声，从而降低任何相关程度。 Armspan还包括胸部宽度，而腿部尺寸似乎反映了单条腿的尺寸。 如果我们有臂长，则可能与腿长有更大的相关性。 但可惜的是，在现实世界中，您很少拥有完美的数据，尤其是当您从互联网上抓取数据时。

So, the correlation between armspan is smaller than expected, but still fairly large. It would be great to test this in larger datasets, so we could look at the effects of leg length on winning chances, if anyone knows of this kind of data, let me know!

因此，臂展之间的相关性小于预期，但仍然相当大。 在更大的数据集中测试这一点将非常棒，因此我们可以看看腿长对获胜机会的影响，如果有人知道这种数据，请告诉我！

翻译自: https://medium.com/@thomasrichierichardson/the-correlation-between-arm-and-leg-length-in-mma-fighters-a-tutorial-on-data-analysis-in-r-3936efe12cb4

mma12注册机