Sample and precision requirements
ANNEX IISupplementary provisions
ANNEX II Sample and precision requirements 1. The Second Adult Education Survey data shall be based on nationally representative probability samples. The weighting factors should be calculated taking into account in particular the probability of selection and the available external data relating to the distribution of the population being surveyed, according to sex, age groups (25-34; 35-54; 55-64), education (at most lower secondary education (ISCED level 1 or lower, ISCED 2 and ISCED 3 programmes of less than 2 years); upper secondary (ISCED levels 3 and 4); tertiary education (levels 5 to 8)); employment (employed or unemployed); region (NUTS II level), where such external data are held to be sufficiently reliable at national level. 2. According to the Annex to Regulation (EC) No 452/2008, the sample size of the Second Adult Education Survey shall be established on the basis of precision requirements that shall not require effective national sample sizes to be no larger than 5 000 individuals, calculated on the assumption of simple random sampling. The absolute margin of error for the indicator mentioned in point 3 shall not exceed the threshold stipulated in point 3, unless this requires an effective national sample size larger than 5 000 individuals. In the latter case, the size of the required effective national sample shall be 5 000 individuals. 3. The Second Adult Education Survey should be designed in such a way that the estimate of the absolute margin of error does not exceed 1,4 percentage points for the estimated participation rate in non-formal education and training for the total reference population aged 25 to 64. The same requirement is relaxed to a threshold of 1,7 percentage points for countries with a population aged 25 to 64 of one million to three and a half million. The same requirement is relaxed to a threshold of 2,0 percentage points for countries with a population aged 25 to 64 of less than one million. 4. The absolute margin of error mentioned in point 3 is the half-length of the 95 % confidence interval. Effective national sample sizes should be adjusted for design effects and anticipated unit non-response to determine the real sample sizes at planning stage. These requirements apply to a sample of resident individuals aged 25 to 64. In the case of national samples with a wider scope, estimates for the resident population aged 25 to 64 should comply with the precision requirements set out above.