How did you calculate the sample size of 2,150?

I recommend you use G*Power software (there is a free version to be downloaded ).

depending on the criteria of your research, hypotheses, and measures, you should select the appropriate statistical test, - correlation, regression, etc., test family, and type of power.

if you define the aforementioned elements properly, you should get a sample size that is much less than what you have and suitable for your research.

here is the download link

https://www.psychologie.hhu.de/arbeitsgruppen/allgemeine-psychologie-und-arbeitspsychologie/gpower

Mahmooda Razzaque -

Are there 59 strata? Have you considered cluster sampling?

Is this a quantitative study? If continuous data or proportions, you use a different 'calculation' accordingly. Online "calculators" generally assume proportions (yes/no data), simple random sampling, No finite population correction factor, the 'worst' case, p=q, which assumes the largest standard error (which is not possible with continuous data), and is therefore often used when it is entirely inappropriate.

Especially if you have continuous data, you need a good guess at the population standard deviation. Actually since you said you used stratified (random?) sampling, you need the standard deviation for each stratum to do a good job. Standard deviations are fixed, and just estimated more accurately with a larger sample.

Methodology and standard deviations are important in arriving at a sample size. You want one which gives you an acceptable standard error for mean or whatever you want to infer for the population. Standard errors become smaller with larger sample sizes.

So it all depends on your methodology, standard deviation(s), and desired standard error. For complex designs, sample size requirements might best be determined by simulation.

FYI, here is a paper where I compared the format for sample size determination for simple random sampling using continuous data to the format for a ratio model under the model-based approach:

https://www.researchgate.net/publication/261947825_Projected_Variance_for_the_Model-based_Classical_Ratio_Estimator_Estimating_Sample_Size_Requirements

Cheers - Jim

Thank you very much Prof.

This is quantitative study, and i am using stratified sampling technique and 59 was total institute in which total population was 9795.

on basis of population (9795) , sample size was calculated by using Krejcie and Morgan,1970) that is 2150.

What i am thinking ( 59 institute) population 9795 is too large to stratify the population. Therefore, I have used another as per Yamane’s (1967) formula:

"n=" "N" /〖"1+N (e)" 〗^"2"

n=59/〖1+59 (0.20)〗^2

n=59/3.36

"n = 18"

Simple random technique was used to select 18 institute out of 59 institute. (code number eg: 001,002 etc.) Manual lottery method.

Now, I have 18 institute and total population is 3130.

I have stratified population (3130) gender wise and year of study wise.

For picking the number of participants in each strata, formula by Cochran’s (1977) applied.

nh "= " Nh/N eg: n_h= 50/3130=0.015× 20=37.8

Thank you very much all Professors.

This is quantitative study, and i am using stratified sampling technique and 59 was total institute in which total population was 9795.

on basis of population (9795) , sample size was calculated by using Krejcie and Morgan,1970) that is 2150.

What i am thinking ( 59 institute) population 9795 is too large to stratify the population. Therefore, I have used another as per Yamane’s (1967) formula:

"n=" "N" /〖"1+N (e)" 〗^"2"

n=59/〖1+59 (0.20)〗^2

n=59/3.36

"n = 18"

Simple random technique was used to select 18 institute out of 59 institute. (code number eg: 001,002 etc.) Manual lottery method.

Now, I have 18 institute and total population is 3130.

I have stratified population (3130) gender wise and year of study wise.

For picking the number of participants in each strata, formula by Cochran’s (1977) applied.

nh "= " Nh/N eg: n_h= 50/3130=0.015× 20=37.8

Yamane is a very narrow/special case of Cochran, both of which are for simple random sampling and for proportions.

Are the institutes different and are they important to you. If so I would have all 59 institutes and plan a multistage sample subsequently analysed by a two level model.

Kelvyn Jones @

Institutions are almost same characteristics which I want to measure.