How to remove 5&95 percent outliers by column independently

Question

Mekala balaji 2017-9-20

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/357568-how-to-remove-5-95-percent-outliers-by-column-independently

回答： Walter Roberson 2017-9-20

Hi,

I have below data, and want to remove outliers by column independently,

vector:

I use the below function, but it can only be used for column array. How to use for the matrix?

percntiles = prctile(vector,[5 95]) 
outlierIndexes = vector < percntiles(1) | vector > percntiles(2);
outliers = vector(outlierIndexes)
nonOutliers = vector(~outlierIndexes)

2 个评论
显示无隐藏无

Walter Roberson 2017-9-20

Do you want to leave "holes" in the array, or do you want to replace the outliers with 0 or with NaN ?

Mekala balaji 2017-9-20

replace outliers with NaN.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2017-9-20

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/357568-how-to-remove-5-95-percent-outliers-by-column-independently#answer_282470

Take the per column low percentile which gets you a row of values. Replicate into as many rows as the original. Compare to the original, getting out a logical matrix. Do the same with the high percentile, getting a second logical matrix. Or the two. Now use that as a mask to assign nan to those locations