I originally posted this in another forum, but in retrospect thought it might be better for this one.
I'm new to this site and seeking help regarding stats for a research project. My stats background is minimal (one college level class) and I would really appreciate any insight you could provide.
I have a sample set generated by a laboratory that consists of 64 samples (32 blinded duplicates). Each sample (n=64) was assayed in triplicate. I'd like to to evaluate how well the laboratory was able to obtain the same answer for the blinded duplicates. The data I have is quantitative. I have previously evaluated this by translating the data into binary form (detected or not) and finding % agreement as well as correlating and getting an R^2 value, but I'm wondering if there's something more sophisticated that I can do to take advantage of the quantitative data. These data are log-transformed to approximate a normal sample distribution.
I have a lots of experience with Excel and some experience with the StatPlus package for Excel (basically the data analysis Excel Add-in for Macs). I also have access to MatLab and R but have very little experience with them, so if there are any built in functions that will help me please let me know! Or, if you can point me to a previous thread
I think a correlation would be fine (you didn't correlate the binary data, right?). Did you look at the plot? I'm not sure what else one could do, but that could just be my ignorance.
No, I didn't correlate the binary data. I looked at the Y-intercepts for systematic bias (none noticeable) and R2s (ranged between .79 and .99-- I have multiple sample sets) and they seem good to me, but there's not really a published acceptance criteria (I'm working with data from environmental microbial source tracking assays) hence my uncertainty. Thanks though!
Reply With Quote
