[R] Remove error data and clustering analysis

[guodong wang]

Hi, all,

I’d like to do the clustering analysis in my dataset. The example data
are as follows:

Dataset 1:

500, 490, 486, 490, 491, 493, 480, 461, 504, 476, 434, 500, 470, 495,
3116, 3142, 12836, 3062, 3091, 3141, 3177, 3150, 3114, 3149;

Dataset 2:

506, 473, 495, 494, 434, 459, 445, 475, 476, 128367, 470, 513, 466,
476,482, 1201, 469, 502;

I had so many datasets like that. Basically, every dataset can
classify one or two clusters (no more than 2), meanwhile, there have
error data points, for example, 12836 is error data point in Dataset
1; and 128367, 1201 is error data points in dataset2.

The clustered data is following the normal distribution, the standard
deviation was known. That’s mean the one cluster is following the
normal distribution when the dataset classified one cluster like
dataset2; the two clusters are following the normal distribution
respectively when the dataset classified two clusters like dataset1.
Error data are far away of the mean.

    I am wondering is there any mathematic pipeline/function can do
the analysis that removing error data, and clustering the dataset in 1
or 2 clusters?

    Thank you for your reply.