Protocol Online logo
Top : New Forum Archives (2009-): : ELISA and Immunoassay

statistical test for ELISA results - (Apr/16/2012 )

I investigated effects of a certain chemical agent on recombinant protein productivity in a CHO cells stable cell line expression erythropoietin. After treatment, I performed ELISA test to investigate positive or negative effects of the agent. I had three group of control cells and three of test cells and ELISA tests were carry out in triplicate for each sample. I am not so familiar with statistics and I do not know what kind of test should I run for my experiment.

here there are my results:

control1: 463 ng/ml
control 2 466 ng/ml
control 3 468 ng/ml

test 1: 600 ng/ml
test 2: 640 ng/ml
test 3: 580 ng/ml

I appreciate any advise.



Use scatter plot with the mean values.


Hi Sottrios
Thank you very much for your comment, but actually I wanted to know the kind of statistical tests like t-test or anova ... to use


Sample 1
Variable : Control
Sample size = 3
Lowest value = 463.0000
Highest value = 468.0000
Arithmetic mean = 465.6667
95% CI for the mean = 459.4151 to 471.9183
Standard deviation = 2.5166
Standard error of the mean = 1.4530
Sample 2
Variable : test
Sample size = 3
Lowest value = 580.0000
Highest value = 640.0000
Arithmetic mean = 606.6667
95% CI for the mean = 530.7750 to 682.5583
Standard deviation = 30.5505
Standard error of the mean = 17.6383
Difference : 141.0000
95% CI : 91.8622 to 190.1378
t=-7.967 DF=4 P = 0.0013

-Ben Lomond-

If each of the three values for the controls and treatments that you present are derived from three separate experiments and each value is the mean of the triplicate assessments, then the analysis presented above is appropriate and you are 99.87 % confident that there is an effect. If the results you present are simply the results of triplicate assessments of the same two samples (test and control) then the analysis presented above is not appropriate.

-Ben Lomond-

Mann whitney U test can be done for small sample size.


you would need a larger set of data to demonstrate normality to establish if the parametric (t-test) if preferable over the non parametric test (u-test). Both will show a clear significant difference between the two groups.

-Ben Lomond-