I am trying to find out a suitable population size for a survey that I want to develop.
I want alpha to be 0.95
The survey will have 40 questions, and each question with have 4 choices.
Thus, my question is, what is the size of the population that I should choose to make sure that I am 95% certain that I believe the results.
Also, so I can do this next time, can you explain the type of statistical test you used, and it's name? Thanks.
ANSWER: In order to estimate the sample size you require some preliminary information mainly regarding the outcome of interest/dependent variable, whether you you are trying to estimate arithmetic mean or proportion(percentage). This information has to be based on some prior experience such as literature review /similar studies or pilot studie if no such information available.
So if you could state your objectives and some related information about main dependent variable, the sample size can be estimated.It seems you are conducting study based on 4 responses.I feel outcome/result can be expressed in terms of percentage, if you are using some scales it can be expressed as mean or median.
Regarding the statistical test for application, it is again based on object of comparison,study design and number of comaprative groups you are having , in relation to indepndent varaibles either parametric or nonparametric approach can be used.But at the moment I cant say which of the methods to be used for your study.If you could specificy your objectives and other methodolgical details, hypothesis etc, then it can be decided about sample size and statistical procedures.
With best wishes.
---------- FOLLOW-UP ----------
QUESTION: Thank yo ufor responding to this. I should have stated before that this is an exteremly non-technical survey, only being used to get feedback for a non-profit. So basically, I don't understand many statistics words at all.
I realize that the larger the sample size, the better the results. I'm just looking for a way to determine this.
Answer If your response variable is in terms of percent, you will get optimum minimum sample size when p =50% =0.50 ,
the sample size reqired will be based on the following formula
n = (Zalpha)^2 *p*(1-p)/d^2
where Zalpha = 1.96 if you want sample size with 95% confidence
d = error tolerable around p,usually it is taken as 10% =0.10, so that your study will show response as 50%+-10% , i.e 40% - 60%
With these condition you require 96 samples. if you decrease value of d, the sample increase.
If you take d=5%=0.05, the required sample size will be 384.
