16 replies to this topic

[ujla80]

hi  

when we pool samples how do we calculate standard deviation and variance?

thank you

[sgt4boston]

"pool samples"...do you mean combine sample volumes into one container?  Or obtain samples and have them as a single population?

[ujla80]

combine sample volume taken from subjects of same study group. suppose we have 8 subjects in one group then combining the sample obtained from each subject to form one sample.

sgt4boston, on 16 December 2010 - 03:23 AM, said:

"pool samples"...do you mean combine sample volumes into one container?  Or obtain samples and have them as a single population?

[sgt4boston]

Never came across this approach.  Usually collect samples from control and experimental groups and compare them (compare the group as a whole or break down into subsets).  

If pooling samples from experimental then you would need to do same with control.

Individual samples should be taken from two groups then analytically compared.

[ujla80]

yes ofcourse we have to treat test and control groups in same way....what if we need 50 microL of sample for a particular sample but we are getting 10 microL from each subject... in this case we pool samples. if we have 8 subjects in a group (test as well as control) we will be having 80 microL sample..than we can use 50 microL sample to perform experiment.The value which we get as a result of the experiment is considered as mean value...so we have two mean values one from test and one from control.So, now the problem is the comparision...we wont have standard deviation (SD) to calculate significance of result...so my problem is that how can we compare two means widout SD?  

sgt4boston, on 16 December 2010 - 01:02 PM, said:

Never came across this approach.  Usually collect samples from control and experimental groups and compare them (compare the group as a whole or break down into subsets).  

If pooling samples from experimental then you would need to do same with control.

Individual samples should be taken from two groups then analytically compared.

[bob1]

Essentially you can't - you need to get more samples(groups) which you can compare to your control.

[Chakchel]

For most experiments not the volume but the concentration of amount of substance matters.
I don't know whether your experiments are exceptions, but if not, then why don't you just fill up your 10 µl sample to 50 µl?

[hobglobin]

I agree with bob1, you neither have a mean nor standard deviation: if you pool your samples to one sample and then measure, you have just one data point.
Chakchel's idea would be a solution (if it's possible), then you have several measurements and can calculate mean + SD.

[ujla80]

it depends upon the sensitivity of method...10µl will have less amount of analyte than 50µl.
for exmaple: if 10µl sample contain 2mg protein then 50µl will contain 10mg.

Chakchel, on 17 December 2010 - 01:40 AM, said:

For most experiments not the volume but the concentration of amount of substance matters.
I don't know whether your experiments are exceptions, but if not, then why don't you just fill up your 10 µl sample to 50 µl?

[ujla80]

sorry...2 means to compare but no SD...

hobglobin, on 17 December 2010 - 02:14 AM, said:

I agree with bob1, you neither have a mean nor standard deviation: if you pool your samples to one sample and then measure, you have just one data point.
Chakchel's idea would be a solution (if it's possible), then you have several measurements and can calculate mean + SD.

[hobglobin]

Well, then two values, one of the treatment and one of the control. But this are still not means, but two single measurements.
You cannot measure a mean, but only calculate it from several measurements, counts etc.

[rkay447]

I agree.  You can not consider the value you obtain from your pooled samples a mean since you don't know the value of each individual sample that made up the pooled.  You are going to have to do the experiment three times and make three independent pooled samples to obtain the needed values for statistics.

[ujla80]

ok... i agree that is the only way we can get valid data.at least three values are required to calculate SD and significance.

rkay447, on 18 December 2010 - 06:22 AM, said:

I agree.  You can not consider the value you obtain from your pooled samples a mean since you don't know the value of each individual sample that made up the pooled.  You are going to have to do the experiment three times and make three independent pooled samples to obtain the needed values for statistics.

[hobglobin]

An idea to make it a bit clearer perhaps. You can calculate mean and standard deviation, if you think your approach out:

You pool the samples in the treatment (and control), let's say 8 samples.
You measure, the result is: 89.0 mg protein in pooled sample
Mean = 89.0 mg/8 = 11.125 mg protein per sample.
Standard deviation = 0 (all samples have the same protein content, because of pooling)

[Gerard]

It's possible to tell if the two samples are statistically different from each other if the the specs ( as VC at the same level) from the method are known.
We had a simular problem and after consulting our biostatistic department we could tell if the groups where different or not.
As I remember one of the big issues was the presence of very abnormal values in 1 one of the samples and the confidence of the test depended strongly of the number of samples in the pool.