I agree with the multinomial regression suggestion. However If you only have few subjects in some age groups you may need to cut down on the number of groups. If age is the only IV and it can take on the 8 group values, I would think that you would want a sample size on the order of 100+ total. Do others have a feeling on this? Best wishes.

Z test for proportions may be used under assumption of normality for large populations.

I think Chi-square test will do that for you

Thank you very much  for all your valuable comments and suggestions.  I apologize in advance if my question is fool. I am taking a course in statistics and I was given a database about HPV prevalence in a population aged 18-25 years old. I have read some articles and the way they presented such data to see if I can perform other analyses.

This article : High-risk and multiple human papillomavirus infections among married women in Can Tho, Viet Nam (http://www.wpro.who.int/wpsar/volumes/03/3/2012.3.1.007_OR_Vu.EN.pdf?ua=1) presents the information regarding infection with high risk/low risk by age group. Clearly, the group of women aged

ANOVA with post hoc analysis may be more appropriate

The problem with proportions that add up to 100% has been noted by Jochen. These form a Dirichlet distribution in which the final proportion is always fixed by the sum of the remainder. 

There is a user-written routine available for Stata here : http://econpapers.repec.org/software/bocbocode/s456725.htm.

There's probably something similar available in R. 

http://econpapers.repec.org/software/bocbocode/s456725.htm

Excellent, Dr. Conroy.

If you want to compare differences in proportions coming form samples with different sizes, you can use the R function prop.test(c(n1,n2),c(N1,N2)) where n1= number of individuals in sample N1 and n2 =number of individuals in sample N2.

Best,

Ovidiu

The Z test for proportions may be used.assuming population size is large n hence normality assumptions may be fulfilled.