Hello Sorry for cross posting.
I was wondering if someone could provide me with advice regarding the way to analyze the following data resulting from a study of insect emergence in a lake. In a small bay of a lake 5 emergence traps were randomly placed in each of 3 habitat types; bog, vegetation and open water for a total of 15 traps. Due to the logistics of funding, collecting the samples, site access and time involved processing the insects only one bay in one lake could be studied. The traps were operated continuously from June until the end of Sept totaling 51 collection dates. Beyond examining the emergence timing and numbers I would like to compare insect emergence from the three habitats at the order level and also at the species level. Since only one bay could be sampled the 5 traps in each habitat are not true replicates (If I considered the 5 traps as replicates then I would have pseudoreplication (Hurlbert 1984), correct??) but rather subsamples so I will use them to calculate a mean # of insects emerging for each habitat. How would I best compare emergence among the three habitats? Could I use a two-way anova without replication using the habitats as the treatments and collection dates as replicates (Sokal and Rolf Example box 11.2). Or is this more of a "repeated testing of the same individual"??? I gather the two-way anova without replication would still apply. Or, could I use a one way anova, habitats as treatments and dates as replicates The next question is if I do use the collection dates as replicates (if it is valid to do so) would I only use the dates during which a particular species emerged? If I use all 51 collection dates then there can be as many as 48 dates of zero counts for all habitats which results in no significance in any test. However, if I only use the dates when a species emerges then there would be a significant difference between the habitats. A twist to this would be what to do with species with one or two distinct generations. Would I only use the dates that the species was collected in or include the intervening zero emergence dates? I have searched and read a number of articles related to such studies and statistics books but either the study is pseudoreplicated or the statistical book examples do not deal with "real world" situations, are obtuse (at least to me) in their explanations or do not even consider "pseudoreplication". I am surprised that in the 20 years since Hurlbert's paper on pseudoreplication, it appears to still occur regularly in ecological studies. Is this because many ecological studies "have to be" pseudoreplicated just from the logistics of data collection and inability to manipulate habitats? Maybe it is because Hurlbert and most other references only superfically dealt with ecological studies (mensurative) and do not provide logistically feasible solutions to dealing with the problem in ecological research. Anyway, hope someone can enlighten me. Thanks in advance dale -- ____________ Dale Parker Homepage (AquaTax Consulting/WWII Aviation) http://www3.sk.sympatico.ca/parkerdw/ Saskatchewan Aquatic Insect Web Page http://www.usask.ca/biology/skabugs/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
