Skip to main content

Table A2.2 The difference between two overall average MPRAB(F) values within a row (Region) in Table 4 is declared significant (based on \(\alpha =0.03\)) if it exceeds 3.70 in absolute value. The primary results of the post-hoc analysis for differences in overall average MPRAB(F) values are listed in this table

From: Sweet tweets! Evaluating a new approach for probability-based sampling of Twitter

1.

For the Recent Method, the average MPAB(F) values are very consistent across Size within Atlanta and Chicago. In Phoenix, Baltimore, and Pittsburgh, the overall average MPRAB(F) values decrease with Size. Notably, the former represent the most populous MSAs in our experiment, whereas the latter represent the least populous MSAs. Thus, the methods consistency was inversely proportional to both MSA population and correspondingly, Twitter volume.

2.

Across all Regions, the average MPRAB(F) values for the Uniform Method were not distinguishable across the different Sizes. The average MPRAB(F) values from Uniform samples were consistent and significantly the highest across Regions and Size with values between 30% and 40%. One exception is that of Size = 360 in Baltimore where Recent and Uniform are indistinguishable. Another exception is in Pittsburgh where for Sizes 360 and 540, Recent performs worse than Uniform but Recent is much better than Uniform for Size = 720.

3.

For Chicago, Atlanta and Phoenix, MPRAB(F) values for Uniform samples, regardless of Size, were significantly larger, on average, than for samples of any Size gathered from the Recent method. Uniform samples of any size in Baltimore produced larger MPRAB(F) values, on average, than Sizes of 540 and 720 from the Recent Method and in Pittsburgh Uniform samples had smaller mean MPRAB(F) values compared to those from Recent samples of Size 360 and 540 but larger than Recent for Size = 720.

4.

VBEST-SRS and VBEST-SYS average MPRAB(F) values were indistinguishable from each other in all Regions for all Sizes and were better than Recent and Uniform in all cases with three exceptions. They were not significantly better than Recent in Baltimore, Phoenix or Pittsburgh for Size = 720.