Hello,

just a simple question, using XmR charts.

If after data plot I have outliers in the mR chart, do I have to remove them from average mR, compute new average mR and control limits for both X and mR charts?

I usually do it this way but I don't know why, this doubt came in my mind recently.

I guess the same approach should be used in X.bar/R Charts.

Thank you for your support.
:thanks:

If after data plot I have outliers in the mR chart, do I have to remove them from average mR, compute new average mR and control limits for both X and mR charts?

I usually do it this way but I don't know why, this doubt came in my mind recently.
:thanks:

Let us use an extreme example. Let's say that the outlier was so huge that it raised the average line (in either chart) such that every other point was below the average line. Is this average line a good predictor of future performance? - No. So, we do remove the outlier and construct the average and control limits without it.

Now, that does leave a quandary. Is the resulting average and control limits a good predictor of future performance? How do we know that the system won't produce another outlier? Assuming the outlier was "bad" we should research what caused the outlier and preferably we either assure ourselves that we don't think it will happen again, or we put countermeasures in place to prevent it from happening again. If the outlier was in the "good" direction, we try to replicate it.

Let us use an extreme example. Let's say that the outlier was so huge that it raised the average line (in either chart) such that every other point was below the average line. Is this average line a good predictor of future performance? - No. So, we do remove the outlier and construct the average and control limits without it.

Now, that does leave a quandary. Is the resulting average and control limits a good predictor of future performance? How do we know that the system won't produce another outlier? Assuming the outlier was "bad" we should research what caused the outlier and preferably we either assure ourselves that we don't think it will happen again, or we put countermeasures in place to prevent it from happening again. If the outlier was in the "good" direction, we try to replicate it.

Thank you Steve for the quick and interesting answer.

You can eventually find the real case with data in my last post ImR and P chart comparison on % defective data (http://elsmar.com/Forums/showthread.php?t=31447)