Hello everybody,
I am doing soil quality study using the attached sampling design. Recently questions are raised regarding the quality of the data collected in that manner.
Originally, the sampling design below shows a hectare plot, divided into 25 plots. The sub-plots for sampling are selected in the form of stratified sampling, with 11 sub-plots. As the plot is only one without any replicate, researchers said the statistical analysis done on the samples will cause complication.
Hence, the sampling design had been changed. In the one hectare area, the sub-plots are changed into plots. In total, there would be 11 plots. From each plots, 3 samples will be collected randomly.
Will the new sampling design produce quality data ? and by using this new sampling design, can i eliminate pseudo-replication ?
You'll need to clarify the question a bit. What question is the study trying to answer? What statistics were you planning on running? If all you want to do is to characterize the soil quality in this plot as a whole then the original design is fine- you can calculate the mean, SD, and SE.
Remember, pseudo-replication is not just a product of study design, it's about analyzing the data from a study in a way that treats non-independent replicates as if they were independent. You might have a well designed study but commit pseudo-replication because you analyzed it incorrectly.
Hello Mr William,
Thank you for your answer. This study will be investigating difference in soil parameters at different land use types, mainly primary forest and oil palm plantation. The parameters that will be measured are bulk density, total carbon, enzymatic activity and more. It is said without replicates, the comparison data would not be functional. Hence, the change in sampling design.
OK, so the main factor you are examining is land-use type, and if I'm reading the original post right you only have one plot for each per land-use treatment. Here's the issue: no matter how you design your sampling within each of the plots, you will still only be able to say that the plots differ but you cannot conclude that that is because of differences in land use. Rather than dividing the plots into subplots and tripling your sampling effort within each of those, what you should do is find three (or more) hectare plots of each land-use type and do 11 randomly selected samples in each then you run the statistics with the following factors:
Land-use
Plot(nested within Land-use)
Ideally the plots you select should be randomly selected from a larger number of potential plots within the geographic range you want the data to characterize. I've attached a paper where we had a similar question- effect of land use on habitat quality- we identified all the areas within our river that met our criteria and then randomly selected 10 for each land-use (shoreline) type.
One final question. In the original post you said the sub-plots were selected based on a stratified design. What are the strata? You'll need to include whatever the strata are as another factor.
Article Effects of anthropogenic shoreline hardening and invasion by...
- Badges
- Science method
- Similar topics
- Mathematical Sciences
- Statistics
- Sampling Design