10.07.2015 Views

KAIS 2007 1 - Kenya National AIDS & STI Control Programme ...

KAIS 2007 1 - Kenya National AIDS & STI Control Programme ...

KAIS 2007 1 - Kenya National AIDS & STI Control Programme ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Normalisation of weightsNormalised weights were used to avoid generating incorrect standard errors and confidenceintervals and were valid for estimation of proportions and means at any aggregation level.They were not valid for estimation of totals, however. Weights were normalised to the <strong>KAIS</strong>sample size and had a mean of 1.0Population estimates (Reference needs to be updated to 2006 projections)Estimation of adult population sizes through extrapolation provided a useful measure of thenumber of adults affected by a particular infection or accessing particular HIV services. Inthis report, we calculated estimates of populations sizes by multiplying weighted parameterestimates from the <strong>2007</strong> <strong>KAIS</strong> (proportions or percentages) by national and provincial‐levelpopulation projections. We used national and provincial population projections from the1999 Analytical Report on Population Projections, Volume VII, KNBS (2002) to determinepopulation sizes at the national and province level, and by sex for analyses requiringestimates for women (e.g. PMTCT). The projections for the <strong>2007</strong> population in the report arebased on findings from the 1999 Census. While a national projection for the number ofwomen and men aged 15‐64 years was not explicitly provided in the report, we were able toconstruct a national projection by summing province‐level projections for <strong>2007</strong> for womenand men aged 15‐64 years as shown in Table A6.e. The report provides projections by fiveyearage category; we summed all categories from 15‐19 years through 60‐64 years to arriveat a total.Due to variations between the Census population distributions by province compared to thethe <strong>2007</strong> <strong>KAIS</strong> population distributions by province, the total sum across provinces (or sex)may not equal the national estimate. The number of HIV‐infected persons summed acrossprovinces, for example, may not equal the number obtained by multiplying the nationalprojected population by the weighted national <strong>KAIS</strong> HIV prevalence estimate. Analystsopted to utilize national level estimates and confidence intervals around national estimateswhenever possible, rather than summing across provinces or sex, because national estimateswere more precise than provincial estimates given the larger sample size. In many analysesby province, population estimates for North Eastern province were not presented becausethe parameter estimate was derived from a sample size too small to generate valid estimates.Similarly, where possible, a “one‐step” process for calculating population estimates wasemployed. To estimate the number of pregnant women who were HIV‐infected, forexample, we multiplied the projected number of women in <strong>Kenya</strong> by the weighted percentof women in the <strong>2007</strong> <strong>KAIS</strong> who were both pregnant and HIV‐infected. In a two‐stepprocess, we would have had to first estimate the number of pregnant women and secondly,the number HIV‐infected among those pregnant. The one‐step process was both simpler andmore appropriate given that the <strong>2007</strong> <strong>KAIS</strong> sampling and weighting design producednationally representative <strong>2007</strong> <strong>KAIS</strong> estimates.<strong>KAIS</strong> <strong>2007</strong> 300

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!