correlation study - Is my sample size too small?

#1
Hi there! i'm a pediatrician working in the neonatal ward, and i've recently finished aquiring data about a study that i'm doing -
i'm trying to find a correlation between the timing of an interventional procedure done in utero (the value being the gestational week) and the neurodevelopmental score (the second value) of those children that were born.
unfortuantely this procedure isnt done vedry often, and the current sample size is 20.

what do you think? is it big enough? also, is linear regression a better test?

thanks
 

hlsmith

Omega Contributor
#2
It depends on how strong the association is! How is the score variable formatted? There are no confounders that need to be controlled for (early birth, SES, use of RT, or anything). You have no control group?
 
#3
i ran the numbers, the correlation coeff. is 0.92 which is pretty strong, right?
and yes, there are plenty of possible confounders. i dont know if a control group is possible in this case, because healthy newborns vary in their stats....
 

hlsmith

Omega Contributor
#4
Yes that is a very high correlation. How is the score variable formatted??? There are many types of correlation procedures and the most commonly used are those for continuous variables, which you may not have. If there was a variable that you wanted to control for you could run a multiple regression model.