# How to test for a statistically significant difference between 3 percentages?

I have samples of stories from three newspapers, all testing for the same proportion of stories with a certain characteristic.

The data I have is: np1=16.34%, np2=14.00%, np3=18.29%. I want to demonstrate whether or not there is a statistically significant difference between the three values.

• Are you asking what test to use? Do you have the actual counts w/ & w/o the characterstic. – gung Aug 30 '15 at 20:44
• I have the counts and am looking for what test to use. I've only done quite basic stats before - enough to know how to do linear regression for some of my other tests but this is beyond me. – Harper Aug 30 '15 at 20:46
• Possible duplicate (up to the "5 categories" rather than "3 categories"). – Glen_b Aug 30 '15 at 23:22

## 1 Answer

You simply need to enter the values with the characteristic and without it into a table of counts. Then you can use the chi-squared test. Under the assumption that there are 10,000 stories from each newspaper, I demonstrate the idea below with R code:

x = rbind(c(1634, 1400, 1829),
c(8366, 8600, 8171) )
as.table(x)
#      A    B    C
# A 1634 1400 1829
# B 8366 8600 8171
chisq.test(x)
#
#  Pearson's Chi-squared test
#
# data:  x
# X-squared = 67.937, df = 2, p-value = 1.769e-15

• AH! Thank you. I thought I was doing something to make it too hard for myself. Simple now! – Harper Aug 30 '15 at 22:53
• You're welcome, @Harper. Why not register your account while you're here? You can find out how in the My Account section of our help center. – gung Aug 30 '15 at 22:56