# what is the minimal sample size for my study?

#### galkim

##### New Member
I am planning a prospective study. I want to see the effect of adding a certain medical device to a doctor's clinic on the amount of referrals that doctor will make. the study plan is as follows -
during a period of one year, 8 different physicians would use the device, but only through a certain period at a time, so that each physician would have some months that he may use the device, and some months that he may not (probably one month with it and one month without it). We would then compare each physician's number of referrals in the period that he was able to use the device, to the number of referrals in the period that he was not able to use it, so that each physician is his/her own control group.
from clinical experience, which was never tested, we expect to see a 20-30% reduction in the amount of referrals. it is important to add that each physician has his/her own referral rate, and it varies considerably among different physicians.

my question -
what is the sample needed for this study in order to get a good statistical power (0.8) and a good p value (<0.05) ?

#### Karabiner

##### TS Contributor
You can use a dependent samples t-test for this. A dependent sample t-test
is essentially a one-sample t-test: the n individual difference values between
condition A and condition B are tested with regard to the null hypothesis:
"the mean of the differences is = 0". Therefore, you can use online devices
such as this here for example

HTH

Karabiner

#### galkim

##### New Member
thanks for the reply, but I cant understand which values to input in that website...

#### obh

##### Well-Known Member
Hi Gal,

I would suggest the following:

Distribution: T, Sample: One sample, power: 0.8

Effect size: what effect size do want the test to be able to identify?
If you are not sure please choose "Medium" (this will choose: Effect type: "Standardized" Effect size:0.5)

https://www.statskingdom.com/sample_size_t_z.html

#### galkim

##### New Member
Hi Gal,

I would suggest the following:

Distribution: T, Sample: One sample, power: 0.8

Effect size: what effect size do want the test to be able to identify?
If you are not sure please choose "Medium" (this will choose: Effect type: "Standardized" Effect size:0.5)

https://www.statskingdom.com/sample_size_t_z.html
wow! thanks!
just one last question - the result is the sample size for the entire study, or just the intervention arm? (in my study, the intervention arm are the number times that the physician will use the medical device)

#### obh

##### Well-Known Member
It is the number of pairs/physicians one pair is one physician before and after