Weighting the Census: Extra Info for Burning Nerds

[If you don’t care too much about the technical stuff, you might skip this section.]

What is the purpose of the random sampling?

  • Individuals who do and do not complete the main survey may differ in certain ways, presenting a challenge for those of us trying to accurately represent the BRC population. This problem is called a self-selection bias and it happens with any survey based on voluntary participation.
  • It is possible to correct the biases due to self-selection if we have access to unbiased estimates of the surveyed population by statistically calibrating the sociodemographic characteristics of a sample to reflect the unbiased values for the population. This is called a weighting procedure.
  • In national surveys, the national census is used as a reference to weight (i.e., adjust) each survey. In our case, the random sample at the gate provides the unbiased assessment of our population.
  • Thus, the accuracy and reliability of the Census results was greatly improved by adding a random sampling procedure.

Comparison of the Weighted and Unweighted Results

The following graphs illustrate the difference between the weighted results (based on the random sampling) and the data from the main (online) survey only.

How to read the graphs:

  • The PINK bars represent the weighted results based on the random sampling.
  • The GRAY bars represent the unadjusted results from the online survey.

CensusBlog2014-04-21-image-4

Comparisons of Variables Used to Adjust the Online Survey

CensusBlog2014-04-21-image-2

CensusBlog2014-04-21-image-6

CensusBlog2014-04-21-image-1

CensusBlog2014-04-21-image-3

Main Highlights from the Comparisons

  • The differences between the results from the random sampling and the online survey show that those who completed the online survey were slightly different from those who abstained: Women and US citizens were more likely to complete online survey, while virgins and individuals younger than 20 years old were less likely.
  • The differences were not extreme, but presented a significant bias in our assessment of the BRC population.

Only the weighted results will be presented in remaining posts of the series, as they are our best estimate of the BRC population.

Written by:
Dominic Beaulieu-Prévost aka Hunter
Edited by:
Tabitha Palmer aka Tabicat
Graphs made by:
David Nelson-Gal aka Scribble

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s