Figure 4 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Comparison of estimates from Pew studies to those from the quasi-quota sampling Pollfish survey (solid circles) and the GSS (open circles). Each point is of one of 33 responses for 12 questions. The Pollfish, GSS, and Pew surveys all yield estimates that are in similar alignment to one another.
Source publication
Background
Researchers have many options for web-based survey data collection, ranging from access to curated probability-based panels, where individuals are selectively invited to join based on their membership in a representative population, to convenience panels, which are open for anyone to join. The mix of respondents available also varies gre...
Contexts in source publication
Context 1
... we adjust the AMT survey by raking (as described in Section 2.1), we find the median absolute difference between the corrected AMT estimates and the GSS/Pew estimates is 8.7 percentage points, and the RMSE is 13.5. Figure A4 in the Appendix shows the full distribution of differences. The statistical adjustment brings the estimates into somewhat better alignment with one another, though the change is not dramatic. ...
Context 2
... accordingly view Pollfish as a middle ground between the extremes we have thus far considered. Figure 4 compares results from the GSS, Pew, and Pollfish surveys on the 12 questions that were asked on all three. As is visually apparent from the plot, estimates from the Pollfish survey are about as well-aligned to Pew as are those from the GSS. ...
Citations
... KnowledgePanel is a high-quality, probability-based panel whose members are recruited through an addressbased sample method utilizing the most recent delivery sequence file of the US Postal Service. A random sample of 7,224 from the approximately 55,000 KnowledgePanel members were offered the opportunity to participate in the survey [18,19]. The KnowledgePanel conducted several quality control measures, and the research team included 2 fake conditions within a list of chronic health conditions to identify and exclude careless or insincere respondents [20]. ...
Purpose
Preference-based summary scores are used to quantify values, differences, and changes in health-related quality of life (HRQoL) that can be used for cost-effectiveness analyses. The PROMIS-Preference (PROPr) measure is a preference-based summary score comprised of 7 PROMIS domains. The PROMIS-16 is a new PROMIS profile instrument. We evaluated the measurement properties of PROPr generated from the widely used PROMIS-29 + 2 compared with the PROMIS-16.
Methods
We performed a secondary analysis of data from an online survey of the general US population, with a longitudinal subsample who reported back pain. The survey included both the PROMIS-16 and the PROMIS-29 + 2 profiles. PROPr scores were calculated from each profile and compared by the distribution of scores, overall mean scores, product-moment correlations with pain measure scores (Oswestry Disability Index, Roland-Morris Disability Questionnaire, Pain Intensity, Interference with Enjoyment of Life, Interference with General Activity Scale, and Graded Chronic Pain Scale), and difference in mean scores in subgroups with 13 chronic health conditions (Cohen’s d).
Results
Of the 4,115 participants in the baseline survey, 1,533 with any reported back pain were invited for the 6-month follow-up survey and 1,256 completed it. At baseline, the overall mean (SD) PROPr score was 0.532 (0.240) from PROMIS-16 and 0.535 (0.250) from PROMIS 29 + 2. At both time points, the correlations of PROPr scores with physical and mental health summary scores from the PROMIS-29 and 4 pain scales were within 0.01 between profiles. Using subgroups with chronic health conditions and comparing between profiles, Cohen’s d estimates of the difference in effect size were small (< 0.2).
Conclusion
PROPr scores from the 16-item PROMIS profile measure are similar to PROPr scores from the longer PROMIS-29 + 2.
