population trends
http://plosone.org/article/info:doi/10.1371/journal.pone.0024708#article1.body1.sec2.p5

We conducted six additional analyses for each data set (i.e., buildings, towers) where we excluded species that had population trends with regional credibility measures that were: 1) low quality, or 2) low to medium quality (Sauer et al. 2011). We also excluded population trends with standard errors (approximated as 0.25 of the 95% credible interval) that were: 3) > 1% per year, 4) > 0.5% per year, or 5) > 0.25% per year. Finally (6), we conducted a linear regression on all species using 1/SE² as a weighting factor. One of these analyses (weighted regression, tower data) indicated a negative relationship between collision vulnerability and population trends (β = -0.19 ± 0.08% per year) and two (credibility 1 or credibility 1&2 excluded, building data) indicated significant positive relationships between vulnerability and population trends (β = 0.43 ± 0.20% and 0.47 ± 0.22%per year), but all relationships were weak (r2 = 0.03 – 0.04) and became nonsignificant if covariates describing migration distance (neotropical vs. short-distance vs. non-migratory) were included in the analysis. (Thanks to Dr. Adam C. Smith, Biostatistician, Environment Canada, for suggesting a weighted regression and prompting us to explore the effect of statistical uncertainty in trend estimates).