Threshold choice and the analysis of protein marking data in long-distance dispersal studies
Article first published online: 17 JUN 2010
© 2010 The Authors. Methods in Ecology and Evolution © 2010 British Ecological Society
Methods in Ecology and Evolution
Volume 2, Issue 1, pages 77–85, January 2011
How to Cite
Sivakoff, F. S., Rosenheim, J. A. and Hagler, J. R. (2011), Threshold choice and the analysis of protein marking data in long-distance dispersal studies. Methods in Ecology and Evolution, 2: 77–85. doi: 10.1111/j.2041-210X.2010.00046.x
- Issue published online: 3 FEB 2011
- Article first published online: 17 JUN 2010
- Received 7 April 2010; accepted 13 May 2010 Handling Editor: Robert P. Freckleton
- decision threshold;
- false positive;
- long-distance dispersal;
- Lygus hesperus;
- protein marking
1. A valuable technique in the study of insect movement is protein marking, a quantitative method where individuals are categorized as marked or unmarked based on the amount of foreign protein detected by an enzyme-linked immunosorbent assay (ELISA).
2. Whether individuals are considered marked or not is dependent on a threshold value chosen by the experimenter. The traditional method of choosing the threshold accepts some risk of false positives, wherein unmarked individuals are misclassified as marked. The error rate associated with this method, adopted from the rubidium marking literature, relies on assumptions violated by most ELISA data.
3. We critically examined the effect of violating these assumptions on the false positive rate. In long-distance dispersal studies where the ratio of unmarked to marked insects is high, false positives can seriously bias estimates of insect movement abilities.
4. Simulations demonstrated that the conventional method for choosing a threshold (i) masks the presence of false positives, (ii) results in a 10-fold higher than expected false positive rate, and (iii) relies on assumptions of normality that are rarely satisfied; non-normality produces further increases in false positive rates.
5. We provide some solutions by introducing a new procedure for choosing a threshold that decreases the incidence of false positives and allows data to be corrected for anticipated rates of false positives. This methodology should enhance researcher confidence in the data generated from dispersal studies using protein marking techniques.