Abbott, Langer Association Surveys FAQ#10

QUESTION:  How is the rate of error calculated?

Each position will have rates of error displayed for three databases: Tax-Exempt/Public Databases, For-Profit Databases, and Survey Participants' Databases.  Compensation levels paid may differ remarkably between these three categories.  Rather than combine these databases, this survey provides each separately to allow the user to make his/her own determinations.  Each database is reported with its own "n"; each has its own rate of error.

Tax-Exempt/Public Databases - Rate of Error

There are two types of rates of error within this general category.

1.        Executive Sources

Data is shown on the PC program using a power curve (polynomial regression analyses).  A standard error is reported based upon the actual data shown in this survey data population.

2.        Non-Executive Sources

Information is principally from government and nonprofit organizations, using composite results of the US OES Occupational Employment and Wage Survey.  This US Government survey reports a Relative Standard Error which is reproduced for the year, area, and job family requested.

For-Profit Databases - Rate of Error

There are two types of rates of error within this general category.

1.        Executive Sources

Data is shown on the PC program using a cubic spline curve through both observed and survey data.  The standard error is calculated only from measurements of variance from predicted mean displayed.

2.        Non-Executive Sources

This information on primarily for-profit company employees is from SalariesReview patented online, interactive salary surveys, and Assessor Series databases.  Because data is collected for individual jobs and not reported by either years of experience, level, or compared to an organizational size dimension (as is the case with Assessor Series presentations), the calculation provides a Standard Deviation measure.

Participant Supplied Data

Standard deviations are calculated via the traditional summing of variances, squaring, dividing by n-1, and taking the square root.  This is a pure conventional standard deviation calculation.

A rate of error will not be shown where more than one source of data exists.  The mixing of OES Relative Standard Error from nonprofit/public sources, regression derived Standard Errors, and SalariesReview's Standard Deviations does not allow the presentation of a meaningful, defensible statistical measure as it might combine three different types of error measure.  If shown, the rate of error is derived as described above from the one data source reported.