Stats on Delta Delta Ct Method, Large Std Dev - (Jul/05/2007 )
Hi,
I'm confused as to what is the standard way of calculating statistics (standard deviation and t-test) from the delta delta ct method. Briefly, when I used the Applied Biosystems guide, they recommend incorporating the standard deviation into the delta ct and delta delta ct calculations. However, when I do this I have some very large deviations that essentially make my data meaningless. The major problem that I see is that while my experimental replicates are nearly identical (same sample + reagents run in three different wells) there is a lot of variability in the Ct values of the subjects within the same group. For example, Subject 1 may of cts of 27.88 and 27.77, while subject 2 will have cts of 29.98 and 30.01, however they are both in the same treatment group.
As a result of the exponential function in this calculation, the deviations are creating a large amount of error.
What was recommended to me was the following: Calculate the delta ct values for each subject individually. Average the delta ct values for one experimental group and use that value to calculate the delta delta ct values for each subject in the next group. Calculate the fold change from each subject, average the values, and calculate the sd and run the t-test from the fold change values. While this makes sense to me, and gives me statistically significant values, I am not sure if it is an acceptable method. Any ideas or suggestions? Thanks.
I normally found it best to do t-test or ANOVA on the ct values directly, as the deltadelta ct changes the distribution of your data. Try out to use the Paffl method, best way using the REST software; which also does the comparison between groups. You can download the software here.
I'm confused as to what is the standard way of calculating statistics (standard deviation and t-test) from the delta delta ct method. Briefly, when I used the Applied Biosystems guide, they recommend incorporating the standard deviation into the delta ct and delta delta ct calculations. However, when I do this I have some very large deviations that essentially make my data meaningless. The major problem that I see is that while my experimental replicates are nearly identical (same sample + reagents run in three different wells) there is a lot of variability in the Ct values of the subjects within the same group. For example, Subject 1 may of cts of 27.88 and 27.77, while subject 2 will have cts of 29.98 and 30.01, however they are both in the same treatment group.
As a result of the exponential function in this calculation, the deviations are creating a large amount of error.
What was recommended to me was the following: Calculate the delta ct values for each subject individually. Average the delta ct values for one experimental group and use that value to calculate the delta delta ct values for each subject in the next group. Calculate the fold change from each subject, average the values, and calculate the sd and run the t-test from the fold change values. While this makes sense to me, and gives me statistically significant values, I am not sure if it is an acceptable method. Any ideas or suggestions? Thanks.
Hi,
Well Ct is the observed value and should be the primary statistical metric of interest, I will send you a paper with a rally good review on stats and real time PCR. How you analyze it will depend on your data, how many replicates, if you know concentrations of template you add or not, etc. Things like t-test assume gaussian distribution and equal variances, but really this is an okay assumption only for large sample sizes-something like 30 samples minimum (central limit theorem if you want to look it up). A Wilcoxon two group test may be something to consider. It is explained fully in the paper I send you now. Don;t have too much fun with your data now!
Vcky
I'm confused as to what is the standard way of calculating statistics (standard deviation and t-test) from the delta delta ct method. Briefly, when I used the Applied Biosystems guide, they recommend incorporating the standard deviation into the delta ct and delta delta ct calculations. However, when I do this I have some very large deviations that essentially make my data meaningless. The major problem that I see is that while my experimental replicates are nearly identical (same sample + reagents run in three different wells) there is a lot of variability in the Ct values of the subjects within the same group. For example, Subject 1 may of cts of 27.88 and 27.77, while subject 2 will have cts of 29.98 and 30.01, however they are both in the same treatment group.
As a result of the exponential function in this calculation, the deviations are creating a large amount of error.
What was recommended to me was the following: Calculate the delta ct values for each subject individually. Average the delta ct values for one experimental group and use that value to calculate the delta delta ct values for each subject in the next group. Calculate the fold change from each subject, average the values, and calculate the sd and run the t-test from the fold change values. While this makes sense to me, and gives me statistically significant values, I am not sure if it is an acceptable method. Any ideas or suggestions? Thanks.
PS you will want to report P values and confidence intervals
and it looks to me that maybe you were having large standard devs because the way you propagated error, did you consider that for an eqn. like y= 2^x the relative error in y = ln2*(error in x), where error is your stdev. but, this way is not proper anyway, jsut may explain your problem? anyway hope the article helps!