Calibration Adjustment for
Nonresponse in Sample Surveys
BERNARDO JOÃO ROTA
Statistics
Örebro Studies in Statistics 8 I ÖREBRO 2016
ÖREBRO STUDIES IN STATISTICS 8 2016
BE
RN
AR
DO J
OÃ
O RO
TA
Ca
libr
ati
on A
dju
stm
en
t f
or N
on
re
sp
on
se i
n S
am
ple S
ur
ve
ys
bernardo rota was born in Quelimane, Mozambique. In 2006 he received his Bachelor degree in Informatics in Maputo, Mozambique. He was then awarded a scholarship to pursue with his studies at master level in Applied Statistics at Örebro University School of Business in Sweden (2008-2010). Following this, in 2012 he was enrolled as a Ph.D student, being financially supported by the Swedish SIDA Foundation - International Science Program. His research interest in estimation in surveys with nonresponse with particular focus in calibration estimators.
Nonresponse is a major threat to valid inference in sample surveys. In the nonresponse scenario, the driver of successful estimation is the efficient use of auxiliary information. Various approaches to accounting for the negative effects of nonresponse are proposed in the literature, with weighting the units in the response being one alternative. This thesis studies weighting for units in the response set by calibration approach. It is here suggested to combine two alternative weighting calibration estimators by means of two-step esti-mation and suggested a variance estimator for the resulting estimator. The thesis motivates the use of sample level auxiliary information for estimation in the first step and the use of population level is suggested for the second step. The thesis also introduces the use of principal components multivariate approach in estimation in surveys with nonresponse.
issn 1651-8608 isbn 978-91-7529-160-4