Dear list members, Birds in agricultural landscapes have been counted annually in a Before-During-After Control-Impact study of possible effects from the construction of a new railway. Obviously, much of the ?design? of this experiment was beyond my control; you cannot direct a billion ? project for the sake of a bird study ? There are six control sites and 13 impact sites of different sizes and shapes (because that?s what the landscape had to offer). All impacted patches of agricultural land were included. Counts were made 2002-2015 with a gap for 2011 and 2012. The construction of the railway became a lengthy process that occurred in various sections along the full 190 km. The sequential steps in the process were: Before, Construction (when the actual building took place), Ready (when the railway was ready but no trains were running) and Traffic (when regular train traffic occurred). Each impact site (patch of agricultural land) was subject for these steps during various years and durations. Also, not all impact sites were surveyed during all the steps. There are issues in the data structure of the count data (some over- and under-dispersion, and high proportion of zero counts), but my main concern is the irregular temporal structure of ?treatments? (variable Status below). The citation marks around the word treatment is because these are not independent treatments but rather repeated measures in a fixed sequence. It would be lovely to ?digest? the full dataset with a Poisson GLMM of some sort, e.g. glmer(response ~ Status + (1|Site), family=poisson), but I have a strong feeling that the requirements of such an analysis are violated by the highly unbalanced design. I?ve also considered a repeated measures ANOVA but I assume the same problem applies there. I would appreciate any comments or advice that could bring the analysis of this dataset further. Thanks in advance! Have a nice day! Adjan Adriaan ?Adjan? de Jong Senior researcher Dept of Wildlife, Fish, and Environmental Studies Swedish University of Agricultural Sciences --- N?r du skickar e-post till SLU s? inneb?r detta att SLU behandlar dina personuppgifter. F?r att l?sa mer om hur detta g?r till, klicka h?r <https://www.slu.se/om-slu/kontakta-slu/personuppgifter/> E-mailing SLU will result in SLU processing your personal data. For more information on how this is done, click here <https://www.slu.se/en/about-slu/contact-slu/personal-data/>
Highly unbalanced design with count data
2 messages · Adriaan De Jong, Thierry Onkelinx
Dear Adriaan, I'd at least add a year random intercept to take into account generic time effects. glmer(response ~ Status + (1|Site) + (1|Year), family=poisson) Such model assumes that a "status" has a constant relative effect which is identical on all sites. One could argue that the length of a disturbance is important too. E.g. the bird will return if the disturbance is short but will stay away when the distrubance is longer. However, I doubt that you have sufficient data to fit such model. Best regards, ir. Thierry Onkelinx Statisticus / Statistician Vlaamse Overheid / Government of Flanders INSTITUUT VOOR NATUUR- EN BOSONDERZOEK / RESEARCH INSTITUTE FOR NATURE AND FOREST Team Biometrie & Kwaliteitszorg / Team Biometrics & Quality Assurance thierry.onkelinx at inbo.be Havenlaan 88 bus 73, 1000 Brussel www.inbo.be /////////////////////////////////////////////////////////////////////////////////////////// To call in the statistician after the experiment is done may be no more than asking him to perform a post-mortem examination: he may be able to say what the experiment died of. ~ Sir Ronald Aylmer Fisher The plural of anecdote is not data. ~ Roger Brinner The combination of some data and an aching desire for an answer does not ensure that a reasonable answer can be extracted from a given body of data. ~ John Tukey /////////////////////////////////////////////////////////////////////////////////////////// <https://www.inbo.be> Op ma 11 feb. 2019 om 10:07 schreef Adriaan De Jong <Adriaan.de.Jong at slu.se
:
Dear list members, Birds in agricultural landscapes have been counted annually in a Before-During-After Control-Impact study of possible effects from the construction of a new railway. Obviously, much of the ?design? of this experiment was beyond my control; you cannot direct a billion ? project for the sake of a bird study ? There are six control sites and 13 impact sites of different sizes and shapes (because that?s what the landscape had to offer). All impacted patches of agricultural land were included. Counts were made 2002-2015 with a gap for 2011 and 2012. The construction of the railway became a lengthy process that occurred in various sections along the full 190 km. The sequential steps in the process were: Before, Construction (when the actual building took place), Ready (when the railway was ready but no trains were running) and Traffic (when regular train traffic occurred). Each impact site (patch of agricultural land) was subject for these steps during various years and durations. Also, not all impact sites were surveyed during all the steps. There are issues in the data structure of the count data (some over- and under-dispersion, and high proportion of zero counts), but my main concern is the irregular temporal structure of ?treatments? (variable Status below). The citation marks around the word treatment is because these are not independent treatments but rather repeated measures in a fixed sequence. It would be lovely to ?digest? the full dataset with a Poisson GLMM of some sort, e.g. glmer(response ~ Status + (1|Site), family=poisson), but I have a strong feeling that the requirements of such an analysis are violated by the highly unbalanced design. I?ve also considered a repeated measures ANOVA but I assume the same problem applies there. I would appreciate any comments or advice that could bring the analysis of this dataset further. Thanks in advance! Have a nice day! Adjan Adriaan ?Adjan? de Jong Senior researcher Dept of Wildlife, Fish, and Environmental Studies Swedish University of Agricultural Sciences --- N?r du skickar e-post till SLU s? inneb?r detta att SLU behandlar dina personuppgifter. F?r att l?sa mer om hur detta g?r till, klicka h?r < https://www.slu.se/om-slu/kontakta-slu/personuppgifter/> E-mailing SLU will result in SLU processing your personal data. For more information on how this is done, click here < https://www.slu.se/en/about-slu/contact-slu/personal-data/> [[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models