Download presentation
Presentation is loading. Please wait.
1
Part 4: Margins of Error Frank Porell
Data and Methodology Part 4: Margins of Error Frank Porell
2
Margins of Error All of the BRFSS and Medicare indicators are estimated from sample data are subject to sampling variation. We have computed margins of error for our estimates and distinguish those for which we are 95% confident that they are different from their respective state average rates. For example, we may report that 60% of older persons in a community reported that they had received a flu shot last year with a margin of error of +/- 5 percentage points (55%-65%). This means that if we repeatedly estimated the rate of flu shots with different samples, the true percentage of older community residents reporting to receive a flu shot should lie within the +/- 5 percentage point margins of error in 95% of those repeated samples.
3
Margins of Error The reported margins of error seem to vary among indicators and among communities. What factors contribute to these differences? Since most of our healthy aging indicators are measured as percentages, the main factor influencing the margins of error are the sample sizes of the estimation samples used to estimate the indicators. The next slide illustrates how margins of error vary with sample size for the same estimate of 50%. Larger estimation samples will produce estimates with smaller margins of error. An estimate of 50% derived from a sample of 96 will have a margin of error of +/-10 percentage points (40%-60%) An estimate of 50% derived from a sample of 2,401 will have a margin of error of only +/-2 percentage points (48%-52%)
4
Margins of Error
5
Margins of Error Points to consider about the margins of error for healthy aging indicators The margins of error for indicators estimated from the CMS data are generally smaller than those estimated from BRFSS data because the community MBSF sample sizes are much larger than BRFSS samples. The margins of error for state estimates of indicators are much smaller than those of community estimates because state estimation samples are naturally much larger than those in any community We distinguish those community estimates that are statistically different from state averages with a “B”, “W”, or “ *” using a simple conservative rule of thumb. We require that the margins of error for the community and state do not overlap for statistical significance.
6
Confidence Intervals Overlap?
Community rate greater than state rate (no overlap) No statistical difference between community and state rate
7
The End
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.