how to calculate the appropriate sample size

ibm9999

New Member
Dear teachers,

I do appreciate if you can help me!

I am a nursing student from Hong Kong, I need to design a study for my year end project.

I want to compare the correlation of rating rated by
1 A and B,
2 A and C,
3 A and D.

to assess which group, (B, C or D) provide ratings have better correlation with group A. number of A,B,C,D is equal.

I just have resoruce to assess around 250 sample (it means no. of sample in group A is 250)

previous study showed correlation between A and B and A and C are around 0.2 to 0.3
but A and D is around 0.4 to 0.6
those previous studies was done in different population
the sample size in previous studies were around 110 to 240

how can I calcuate an appropriate sample size?
should I use T test or Z test or other test?

May

jamesmartinn

Member
To determine the 'appropriate' sample size, you need to set the level of Power you wish to achieve. (ie., the probability your study will find an effect) Usually, this is set at .80 or 80%. If you download the program G*Power, it will compute the necessary sample size for you automatically. You can download it for free here: http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/

It will allow you to calculate sample size for a variety of tests (F, t, z, r)

If you are wishing to compare correlations between groups, you can have a look at the following website. (Scroll down to section 4)
http://www2.hawaii.edu/~donaldp/Biostatitics/lecture23.htm

Last edited:

ibm9999

New Member
Thanks Jamesmartinn

I tried to use the G power. but I am not sure how to set the effect size.
from the previous studies, the correlations between groups ranged from 0.1 to 0.6

which number should I use?

In this case, should I use T test or Z test or other test? as it affect the no. of sample size

Thank you very much!

May

frailjar7

New Member
Dear teachers,

I just have resource to assess around 250 sample (it means no. of sample in group A is 250)

previous study showed correlation between A and B and A and C are around 0.2 to 0.3
but A and D is around 0.4 to 0.6
those previous studies was done in different population
the sample size in previous studies were around 110 to 240

how can I calcuate an appropriate sample size?
should I use T test or Z test or other test?

May
Mary, First off, since your sample size, n = 250 a z CI should be used because n is greater than or equal to 30 (250 > 50). To find the n the formula is: X-bar +/- z-sub alpha/2(sigma/sqrt(n)) and the mean, mu is bounded by this equation (on both sides of mu) and a CI table can be found in any Elementary Statistics textbook. And if I may, I would like to suggest one, as a reference, Elementary Statistics: A Step By Step Approach, 6th ed. by Allen G. Bluman (ISBN:978-0-07-304825-3). This is a great reference, I am not a graduate student, but have used this book in undergrad and loved it! This text also includes a laminated table of important formulas and tables for Student's t, the tails, Chi-square, and CVs for PPMC (Pearson Product Moment Correlation). Hope this is not to late. Good luck!