Research Project Pseudoreplication Problem

#1
Hello,

I posted a few weeks ago but some new problems have come up and I could really use some advice! I am joining in on a research project in which the heavy metal content of mussels is being examined at 3 locations over 4 years. At each location, there would ideally be 4 sub-samples of around 10 mussels, for a total of 40 mussels. The 10 mussels in each sub-sample are then homogenized for heavy metal analysis. However, I just found out that in the last 2 years of the project, the mussels from each location have all been homogenized together and then divided into 4 replicates, rather than split up before homogenizing. This is now a case of "pseudoreplication," as there is technically only 1 sample with no variance for these 2 years.

What I am trying to find out now is: what would be the best way to analyze for spatial and temporal trends now that I technically only have 1 sample (40 mussels all homogenized into 1) for each location over the 4 years?

Thanks so much for taking the time.
 
#2
Hello,

I posted a few weeks ago but some new problems have come up and I could really use some advice! I am joining in on a research project in which the heavy metal content of mussels is being examined at 3 locations over 4 years. At each location, there would ideally be 4 sub-samples of around 10 mussels, for a total of 40 mussels. The 10 mussels in each sub-sample are then homogenized for heavy metal analysis. However, I just found out that in the last 2 years of the project, the mussels from each location have all been homogenized together and then divided into 4 replicates, rather than split up before homogenizing. This is now a case of "pseudoreplication," as there is technically only 1 sample with no variance for these 2 years.

What I am trying to find out now is: what would be the best way to analyze for spatial and temporal trends now that I technically only have 1 sample (40 mussels all homogenized into 1) for each location over the 4 years?

Thanks so much for taking the time.
I'm sorry to hear this nikhil, but it sounds like the data has been reduced to 3 samples per year (one for each location). It sounds like there really isn't much you can do - if the differences in concentrations are large enough (and I mean very large) you may still find an effect but the power will be very low. If you can ensure that the procedure is done correctly from now on you may be able to salvage some of the power. For instance, you can do this by treating the 4 sub samples in the first two years as latent variables (i.e. missing data), when you can more data from future measurements - though it sounds like you still wont have much degrees of freedom.