Can we have more information on the population size of patients? Also, how was the sample selection procedure for n=5, before one can provide any reply.

Irrespective of how the sample size n was obtained, the standard deviation being a measure of dispersion is necessary. It gives an idea of the variability amongst the sample items.

What can we do with the measures of variability, when the selected sample is not a probability selection and the sample is not representing the population size.

Hello Ranbir Singh,

Assuming that it is a random sample (H. Öztaş Ayhan 's point), a standard deviation provides some information. But perhaps the question underneath is rather:

How reliable is the SD with n = 5? It always depends on the population being sampled, but my blind guess would "not very reliable". We can do a simple demonstration by simulating samples (n=5) from a normal distribution with known variance and calculating the SD.

The following R code takes 10 random values from a normal distribution with a mean = 0 and sd = 1, then calculates SD in each sample.

set.seed(123)

for (i in 1:10) print(sd(rnorm(n = 5, mean = 0, sd = 1)))

These are the SD's from the 10 samples (n = 5 for each)

[1] 0.8110218

[1] 1.16349

[1] 0.6394632

[1] 1.411041

[1] 0.3445583

[1] 1.261993

[1] 0.5074036

[1] 0.4935599

[1] 1.414066

[1] 0.692965

We can see that the SD's vary wildly! So if providing the SD, it might be honest to mention this caveat...

It depends on the type of data and analysis, but sure, why not, there are no rules :)

There is no rule to do/not. However, the nature of the data you have may lead you to decide which measure of variability/location to use that makes sense for your report.

Yes. Because a sample is the representation of the population, so population parameters must be measured.