Hi Rafael,
I would suggest you to perform Poisson regression for each species separately, cosidering the following model,
log(count) = season + replication
Here season should be considered as fixed factor whereas replication should be random factor. Now while fitting this model you may encounter the problem of zero count. You can get rid of this problem in either of the following two ways:
1. Add an fuzz value near to zero (say, 0.01) to all the cell count.
2. Take average count over all the replication in the model and remove replication from the model. After fitting this model still you may need to add fuzz value.
If seasonal effect comes to be significant then one can tell that fish count of that particular species is affected by seasonal variation.
Now after fitting the
model for each species, you can calculate LSM (least square mean) of count for each season, and then put them graphically plotting count in the Y-axis and seasons in X axis. This will give you clear idea counts of different species over all the season.
I don't find any sense in analyzing all the species data at a time.
Hope this helps.
Regards
Madan
Rafael Rafael <rafaelcabrera72@...> wrote:
Hello,I'm not an expert on ANOVA test and would like to receive helpto solve one problem. I will try to expose my doubts aboutthe test.I have a data set with specimens number (counts) of each 26 fish speciescaught by artisanal fishery on four season (summer, fall,....)My table is:Summer Fall................. ......... .....Winter. ....... Replicate Sp1 Sp2 Sp3 Sp4.....Sp1 Sp2 Sp3 Sp4........... 1 34 11 9 112 11 28 43 145.......... 2 27 17 15 136 6 15 21 89............ 3 33 21 12 109 0 19 33 115.......... 4.......5....... 0 0 06.......7...... 0 08......My doubt:What is the corret methods to test differences in the seasonal species composition ?I have some zero in the table. The species is absent in some replicates and in some seasons. Is there some problem with this zero in the table ?Any help to solve the problem to find the correct test for these data will be welcome.Thank you in advance for your attention to my request.All the best,RafaelSão Paulo - Brazil
Abra sua conta no Yahoo! Mail, o único sem limite de espaço para armazenamento!
--------------
Madan Gopal Kundu
Biostatistician I, i3 Statprobe
Biostatistician I, i3 Statprobe
Gurgaon, Haryana
India
mobile: 91-9868788406
e-mail: Madan.Kundu@...
Click to join Statisticians_group
Unlimited freedom, unlimited storage. Get it now
Offline 
Send Email