Help using Base SAS procedures

How to get reference ranges

Reply
Super Contributor
Posts: 401

How to get reference ranges

Hi, is there a procedure of code in SAS where I can get reference ranges?
For example I have 20 cities with home prices for each month for 2 years for each city.

I do have the standard Deviation and the Coefficient of Variation, but am looking to see if a range of the Prices are more in an optimal range then others.. almost like identifying outliers.  So for example if the range of City-A is between 300,000 and 500,00, is optimal, anyhting outside of that range is considered an outlier, or whatever term you may call it.   Thanks

Super User
Posts: 3,256

Re: How to get reference ranges

Try the UNIVARIATE procedure. It identifies your lowest and highest values plus percentiles, means, medians etc. You could use a BY statement on your city variable to look at percentile variations between cities.

Trusted Advisor
Posts: 2,116

Re: How to get reference ranges

Typically a 'reference range' is an external number that is not derived from the data.

Respected Advisor
Posts: 2,655

Re: How to get reference ranges

Agreed, Doc.  But how about generating that reference range for the 'first' time?  Lately, I have been tasked with doing something like that with historical control data in tox studies.  UNIVARIATE is my tool of choice--normality testing, percentiles, etc. Ultimately, I want to incorporate this info as a prior distribution for some Bayesian analyses.

Steve Denham

Trusted Advisor
Posts: 2,116

Re: How to get reference ranges

Posted in reply to SteveDenham

Steve,

Often the "reference range' comes from historical controls.  Sometimes from specific analysis of known normals (typical for clinical lab tests).  I worry about podarum's approach of using the 'study data' to derive the ranges, particularly for housing prices (we know they have been problematic over the last several years).

Doc

Respected Advisor
Posts: 2,655

Re: How to get reference ranges

Doc,

I agree about using 'study data' for ranges--it's like using EDA methods as confirmatory stats.  Of course you find something, you just spent 15 hours digging around looking for anything shiny.

Steve Denham

Ask a Question
Discussion stats
  • 5 replies
  • 241 views
  • 0 likes
  • 4 in conversation