Paired Vs unpaired t-test - which one to use?

#1
Good morning,

I am a little stuck - I understand the difference between the two tests and am happy with written examples in a stats book I found. However I am still not sure which to use for my data. I am testing two reagents: I have run an assay with all variables the same except the reagent I am testing. Plate 1 contains reagent 1 (samples 1-8) and plate 2 contains reagent 2 (samples 1-8). This assay has been repeated by another operator to discount operator differences. I have analysed the data separately (i.e. operator 1, then operator2) and together both using a paired ttest and a 2 sample ttest with contrasting results.
Operator 1: Paired ttest = 0.166
Operator 2: Paired ttest = 0.049
2: 2 sample ttest = 0.085
All data: Paired ttest = 0.019
2 sample ttest = 0.246

So therefore my questions are:

1) is it actually possible to pair this data as I believe it is? I know the two reagents are on different plates, but previous analysis on the plates has provided consistent results so plate variation wouldne be an issue here.
2) Which would be better for me to use?
The problem I have is I know the answer I want so I am trying hard not to bias my decision. I have a total of 16 samples from both operators and the values range from 3.18 - 3.68 for reagent 1 and 3.22 - 3.71 for reagent 2.

Any help would be gratefully received! I hope my query makes sense!

Regards

Emma
 

Miner

TS Contributor
#2
The paired t-test is for dependent measures while the unpaired (2-sample) t-test is for independent measures.

Is sample 1 on plate 1 the same as sample 1 on plate 2? And samples 2 - 8 the same?

When in doubt, plot the measures against each other. If they are dependent you should get an elliptical plot.