# How do I know an outlier in a time series isn't the beginning of a trend?

I wish to average detections coming in over time.

I use the interquartile range to identify outliers and to discard them. I look at the last 30 values.

What do I do if each new value is an outlier, and hence is discarded? How do I detect that the average has significantly shifted?

I asked a fuller version of this question on stats.stackexchange.com.

-