Question

1. Describe some ways to identify extreme score(either visually or z-score). Discuss why it is desirable to identify the extreme score in a distribution.

2. You are reading an article in a journal that describes research based on data with poor reliability. The author argue that reliability of 0.70 is common and acceptable. Do you agree with this position? If not, how would you counter that commonly held position.

Answer 1

1) few ways to identify the extreme scores include box-plot, Z-score, top and bottom 2-5 %ile. It is important to identify the extreme values as they may deteriorate the measure such as mean if one or few of the values are very high which would wrongly pull the average upwards brining in the bias.

2) Reliability of 0.70 may or not be enough. For instance, if healthcare, where life is at stake, 0.70 reliability won't be appropriate. However, in retail campaign, it may be fine to have such reliability. So, it depends on industry and nature of the problem that one is addressing.