Outlier and Pearson corellation of coefficient

M

megaproto

New member
Joined
Jul 9, 2011
Messages
4
hello let say there are 2 set, food and weight of hippopotamus, with one set (the weight) having one outlier (for discussion let say one of the hippo weight 1000 KG which is considered an outlier in comparison to other data). The hippo is given food A and after that the weight were recorded. Later the hippo is given food B and the weight were then recorded. For food A the hippo with outlier data increase in weight to 5000 KG, while the other hippo are around 200 to 300 KG. For food B that hippo increase in weight to 5500 KG while others are around 220 to 320 KG. In order to look at the relationship between the weight of food A with normal diet and food B with normal diet, should I compare the median value rather than the mean value of the weight for pearson corellation of coefficient?

thanks :)

one more question about statistic, if the data is not normally distributed which mean median and average is not the same (right?) but there is no outlier, i should use mean correct? for Pearson corellation of coefficient

thanks :)
 
If you suspect an outlier, you are better off with the median, independent of other distributional assumptions. One outlier can severely distort the correlation coefficient. This is one reason you should always graph your data in some form or fashion. However, you cannot compute a reliable Pearson correlation with only two data points, since the answer can only be 1,-1, or undefined.
 
megaproto said:
hello let say there are 2 set, food and weight of hippopotamus, with one set (the weight) having one outlier (for discussion let say one of the hippo weight 1000 KG which is considered an outlier in comparison to other data). The hippo is given food A and after that the weight were recorded. Later the hippo is given food B and the weight were then recorded. For food A the hippo with outlier data increase in weight to 5000 KG, while the other hippo are around 200 to 300 KG. For food B that hippo increase in weight to 5500 KG while others are around 220 to 320 KG. In order to look at the relationship between the weight of food A with normal diet and food B with normal diet, should I compare the median value rather than the mean value of the weight for pearson corellation of coefficient?

thanks :)

one more question about statistic, if the data is not normally distributed which mean median and average is not the same (right?) but there is no outlier, i should use mean correct? for Pearson corellation of coefficient

thanks :)
Click to expand...
The Spearman rank correlation is also can usable in the same cases because,it can handle the outliers. The Spearman and the Pearson gives you different results. You can compare this results in this calculator:

http://www.coreminer.com/calculators/correlation-test
 
You must log in or register to reply here.
Top