# Graph with error bars - how do I do them? - (Apr/08/2008 )

Dear All,

I have been doing a time course experiment (analysed by Real Time PCR) and I have repeated it 3 times (independently), obtaining 3 different and independent sets of results for the same time course.

I want to represent it in a graph with columns and error bars, but I am not too sure of how to do it and what statistical commands to use in Excel.

As far as I know, the average between the 3 values for a certain time point would be used for the column, right? But how about the error bars? Do we use standard deviation for that? Or standard error? SEM? Or something else?

If we are supposed to use standard deviation, then which type should it be? I noticed in Excel there are at least 4 different ways of finding standard deviation (STDEVP, STDEV, etc)

I would really appreciate if you could help me out here, I've checked some tutorials on the internet but I still couldn't figure it out.

Thank you very much!

Julianne.

-Julianne W-

Hi Julliane,

a colleague of mine had the same question last week. I would say that you should use the standard error with your error bars. My reasoning for this is that you are trying to present the average value of a population of many (infinite) measurements at a certain time point (that would mean if you did let's say a million measurements at that time point, you would be quite sure that the average of those results is the actual value at that time point) by approximating it with your average of three measurements and using the error bars to represent uncertainty. And when doing that the standard deviation of a sample mean is not quite the right statistic to represent dispersion, so we use the standard error, which is s/sqrt(N), where s is the standard deviation and N is the size of the sample (in your case 3).

I hope my reasoning is right,
regards.
Miha

-BioMiha-

Hi, BioMiha!

Thank you so much for your answer, it is much more clear now! But I was checking excel and there is actually a couple of different commands to find SD. The ones that looked more right to me are STDEV and STDEVP. Do you remember which one you usually use?

Thanks a lot!!

All the best,

Julianne.

QUOTE (BioMiha @ Apr 8 2008, 02:34 AM)
Hi Julliane,

a colleague of mine had the same question last week. I would say that you should use the standard error with your error bars. My reasoning for this is that you are trying to present the average value of a population of many (infinite) measurements at a certain time point (that would mean if you did let's say a million measurements at that time point, you would be quite sure that the average of those results is the actual value at that time point) by approximating it with your average of three measurements and using the error bars to represent uncertainty. And when doing that the standard deviation of a sample mean is not quite the right statistic to represent dispersion, so we use the standard error, which is s/sqrt(N), where s is the standard deviation and N is the size of the sample (in your case 3).

I hope my reasoning is right,
regards.
Miha

-Julianne W-

Julliane

use STDEV and divide it by sqrt (N). STDEVP calculates the s.d. of a population and takes into account that the data you provided are from a population and not from a sample.

Cheers,
Miha

-BioMiha-