Can One Apply A Summary Statistic On Rank Data Of Microarray Gene Expression

Question

[elb83]

Hi guys, I have a list of 6 probes representing the same Gene_Symbol as in a microarray gene expression experiment. The numerical value for each probe is a ranking instead of a gene expression value as normally is used to do because of as output of another analysis. Now, I would like to have a single value representing the single gene. Is it possible to apply the median or mean to such ranking? My doubt is: since this numbers represent a ranking and not a gene expression value, probably i suppose to manage in a different way such type of number. Can anyone help me please to clear such doubt? Thank you a lot!!!!

Answer 1

[Istvan Albert]

The term that describes your scale is Ordinal scale

The link above describes some of the measures that are valid for such a scale. For example median makes sense but mean or standard deviation do not.

The main problem with this scale is that is does not capture the magnitude of differences between consecutive items. Thus the distance from position 1 to 2 may be substantially different than the distance from position 2 to 3.

Further searches using this terminology will also turn up valuable examples.

In general you should look to make use of contingency and frequency tables.

[Istvan Albert]

Found a neat tutorial: http://learnandteachstatistics.wordpress.com/2013/07/08/ordinal/

[elb83]

Thank you lot Dott Istvan Albert. Your help is always precious! thank you again.