I have 20 random points (coordinates) and for each point I have a classification of the dominant specie (only one specie per point), I have 6 species in total. I have 4 years of data (1976, 1985, 1993 and 2018). Therefore for each point I have one classification of specie for each year. Thus, I basically have categorical dependent variables (the species). For each categorical dependent variable, I have two categories (present or absence), which are mutually exclusive. I also have one independent variable (years) with 4 categories (1976, 1985, 1993 and 2018), which are related. I want know if the proportion of the species who were present or absent differed in each year and which species and which year differed. My data is not normal neither present homocedasticity (even after transformations). Is it appropriate to perform Cochran’s Q test and post hoc McNemar? I have tried some post hoc, but they only one that worked out in R was one using this command from rcompanion package (pairwiseMcnemar(x=mydata.long$presence, g=mydata.long$oitocinco, block=mydata.long$id, method="bonferroni", test="mcnemar"). It seems it is a post-hoc McNemar approximate using chi-square distribution (no continuity correction). But I do not understand what this means, neither if it is appropriate. I heard about perform repeated Generalized Linear models (GLM), however I could not find a scrip for it neither how to organise a table for R. Could anyone help me with that, please?
This is too complicated a model to fit into Cochran's Q test.
But if you were analyzing each year entirely separately, you could use Cochran's Q test. You can use the example at the following webpage, substituting Species for Practice and Site for Student. I think the rest of the plotting and post-hoc testing will then make sense.
http://rcompanion.org/handbook/H_07.html
If you want to include year, you would probably need to jump to something like a multinomial logistic regression with Site acting as a random effect. That may be more complicated than you want.
Perhaps to look at changes over years, you might just condense your data to proportions of each each Species overall, and look at these over the years.
Hi Ana
Reading your question above, five related problems have been notified and need to be studied carefully.
A. In order to find the possibility of develop any statistical measures, the following general remarks need to be determined:
1. How many random variables in your research. How many dependent and independent random variables.
2. What is the relation between the random variables, i.e. related or not.
4. What is the sample size of your random variables >30 or
Please refer to:
1. Landau, S. and Everitt, B. S.(2004). A Handbook of Statistical analyses using SPSS.
2.Howitt, D. and Cramer, D.(2008). Introduction to SPSS.
Regards,
Zuhair
- Badges
- Science topic
- Similar topics
- Econ
- Manufacturing
- Industry