How to estimate data repeatability?

Question

[Andrew Liu]

Here I have two sets of replication of the same data, and I want to test the repeatability of this experiment, which value should I pick to illustrate it. I tried Pearson corr, but it can only tell you that your data are correlated but cannot tell the slope of your data. If your data can fit y=2x, it will be correlated but I want to test if it can fit y=x. Thank you !

[rpolicastro]

Can you describe what your data is in more detail?

Answer 1

[Kanne]

I'd always scatterplot the 2 replicates against one another before considering using a metric to summarise the relationship. Perhaps try this first?

After that, if you're most interested in how well the data fits x=y, I would play with something like taking the difference between x and y for each datapoint (yi - xi) and plot these as a histogram.

The shape of this distribution might give you an idea of what to use. For example, the spread of this distribution (standard deviation) could be one metric of how similar the datasets are.

Of course whether this is the right metric for you depends on what your data looks like, how much you want your metric to be skewed by outliers, etc...

[Andrew Liu]

Great, I've tried this. Thanks a lot