I guess it depends on the strata that you're standardising by. Given the formula for direct standardisation

O*= (Sumi(Pi * Oi)) / Sumi(Pi)

with O* the standardised outcome, Pi the size of the standard population in stratum i, and Oi the outcome to be standardised within stratum i (e.g. of a study sample),

the only difference between age-standardisation and age-sex-standardisation would be that in the former, your strata are defined only by age groups, and in the latter, they are defined by a combination of age group and sex. So it depends on what strata your standard population was divided into.

As far as I understand, standardisation on a given characteristic (age, sex, ...) obviates comparisons on that characteristic because you're aggregating across the different values the characteristic can take

HTH, Hans

I agree with Hans. Standardization basically depends on the objective of the analyses and the population structure of the groups that you would like to compare. For example, if one is interested in comparing an aggregate quantity of interest (let us say mortality) between two populations and age is the only (or the primary) factor affecting that process then a simple age-standardization would enable us to investigate the differences in the aggregate quantity of interest, net of the ‘distortions’ that are due to difference in age structure between the two populations.  On the other hand, if we have a reason to believe that the process under study also varies by sex and that the two populations have distinct age and sex distributions then we could undertake age-sex standardization.  We can also perform higher order standardizations, if we further assume that other factors other than age and sex also affect the process of interest.   So, the key point is that first we need to identify the factor (or factors) that affect the process we are trying to study. Secondly, and more importantly, we should have a reason to believe that the groups whose aggregate measure we would like to compare have population structure that is different from each other.  For example, if we are comparing the intensity of mortality between two populations and the two populations have fairly comparable age and sex distributions, age-sex standardization would make no difference as we can directly use the reported crude death rates as they are to compare the mortality differences between the populations.  Here, are a few references that may be of interest: Das Gupta, P. 1978. "A General Method of Decomposing a Difference Between Two Rates into Several Components," Demography 15 (1): 99-111. Or for more recent and general formulation on standardization see Das Gupta, 1993. ‘Standardization and Decomposition of Rates: A User’s Manual, Special Studies, U.S. Department of Commerce.

Thank you Hans and Yohannes for your explanations. I was about to ask the same question when I found Glenn's here. 

Hi Hans Bauer

I am here new, I am working on educational disparities among districts (36) of Pakistan by using MICS survey data (with 250,000 individuals and 38000 households' information).

Please help me in finding age and gender wise standardized education index (for level of achievement) and gini-coefficient (for level of disparties).

Suggest me either to use STATA or SPSS, further i am interested to use INDEPTH standardization.