Various Customer Surveys

12/02/2010

Second Revision of the

Sample Plan for the 2010 Census Partnership Survey

Proportional Allocation

Universe size: ~247,000 Partners listed in the Integrated Partner Contact Database

Sample Size: 7,502 = 1340 * 5 + 802

Strata: 6 Strata Formed

National Partners (Self-representing Strata)	N = 802 n = 802 Base Weight = 1.000
State/Local/Tribal Government Agency Partner Strata	N = 38,716 n = 1340 Base Weight = 28.8925 SE(0.50) = 0.02 90% CI: 0.50 +/- 0.03
Community Organization Partner Strata 9 Strata based on Hard-to-Count and Value Added
Hard-to-Count	Value Added
Low n = 2,680 SE(0.50) = 0.01 90% CI: 0.50 +/- 0.02	$0 N = 36,946 n = 1,340 Base Weight = 27.5716 SE(0.50) = 0.02 90% CI: 0.50 +/- 0.03	>$0 N = 61,950 n = 1,340 Base Weight = 46.2313 SE(0.50) = 0.02 90% CI: 0.50 +/- 0.03
High n = 2,680 SE(0.50) = 0.01 90% CI: 0.50 +/- 0.02	$0 N = 42,526 n = 1,340 Base Weight = 31.7358 SE(0.50) = 0.02 90% CI: 0.50 +/- 0.03	>$0 N = 66,255 n = 1,340 Base Weight = 49.4440 SE(0.50) = 0.02 90% CI: 0.50 +/- 0.03

• Sample distributed proportionally to nonself-representing strata within Hard-to-Count Stratum.

• Sort within State/Local/Tribal stratum by Region, State, State Government Organizations, Regional Organizations, Tribal Organizations, Value Added, Member Count

• Sort within Hard-to-Count by Value Added stratum by Region, State, Partner Type, Number of Members

• This sampling plan will buy us:

  • Coverage of Partners in the Hard-to-Count score, Value Added score, region, and size of membership.

  • National Level Estimates will have a 90% confidence interval of +/- 1%.

  • Within each Hard-to-Count stratum we will have a 90% confidence interval of +/- 2%.

  • Within each Hard-to-Count by Value Added stratum we will have a 90% confidence interval of +/- 3%.

  • We will be able to detect differences in estimates of 4.26% (i.e. if the difference is more than 4.26% then the difference is significant.) at the Hard-to-Count stratum level.

Assumptions:

The assumptions made for this sampling plan:

• Initial response rate is 30% within each stratum.

• We will conduct a nonresponse follow-up of all NR cases with an assumption of 62% response.

These two estimates are used in determining the sample size.

Weighting: Base weight = Inverse of Probability of Selection

Nonresponse Adjustment: Inverse of propensity to response (based on logistic regression)

Second-stage Adjustment to Universe Totals

Variance: Implement a Stratified Jackknife Procedure