How outliers can effect the correlation results (Spearman and Pearson)?

I have two series of data. The source of the data are different but both are temperature that are measured by different instruments with different accuracy.

Please possibly give answers with reference.

  • 2
    $\begingroup$ Check the image in the "Correlation and linearity" section of the wikipedia correlation article. $\endgroup$
    – naught101
    Commented Sep 10, 2013 at 7:09
  • $\begingroup$ @naught101, How will the type of error can effect? I mean being random or systematic. $\endgroup$
    – ali amidi
    Commented Sep 10, 2013 at 7:13
  • $\begingroup$ Not entirely sure what your question in your comment means (grammar is incorrect), but I guess the answer will depend on what the source of the outliers is. For example, are the outliers due to human error in transcription, measurement errors in instrumentation, or just highly-unexpected extremes? The latter are particularly difficult to deal with, because they may indicate your model is wrong (e.g. there is a non-linear relationship, and correlation isn't an appropriate measure). Also, it can be very difficult to separate the sources or error. $\endgroup$
    – naught101
    Commented Sep 10, 2013 at 7:18
  • 2
    $\begingroup$ Pearson is very sensitive to outliers; Spearman isn't, but is still sensitive to the shape of the cloud. See also. $\endgroup$
    – ttnphns
    Commented Sep 10, 2013 at 7:57
  • $\begingroup$ For the standard examples see our posts that reference Anscombe's quartet $\endgroup$
    – whuber
    Commented Sep 14, 2022 at 16:16