Per Gösta Andersson Professor

Professor i statistik

Nedan följer några förslag på projekt för doktorander:

Possible research project for graduate students:

Within survey sampling nonresponse is a big problem. We construct a sampling
design and perform the study but fail to get answers from all selected
participants (unit nonresponse). Calibration estimators have been used for a
long time in order to estimate population totals or means and they can also be
modified to take into account nonresponse, see e.g. Särndal and Lundström
(2005). However there are many possible ways to calibrate and according to
Deville and Särndal (1992), calibration can be linked to a distance measure.
The so called optimal calibration estimator is connected to a distance measure
as discussed in Andersson and Thorburn (2005) and this was followed up
for the nonresponse siuation in Andersson (2019). This latter article presents
point estimators and the current research by myself concerns variance estimation.
To understand how this should be done we go back to Särndal et al.
(2005) and study the linearisation technique they have utilised.
Another issue is how to use all available data we actually have access to
in our time for a specific sampling situation. We would like to have few
efficient auxiliary variables, but how should these be constructed and by
what principles? This, and the previous problem about variance estimation,
are possible research projects for a doctoral student.

References

Andersson, P.G. (2019). ”Design-based ”optimal” calibration weights under
unit nonresponse in survey sampling”, Survey Methodology, 45(3), 533-542.
Andersson, P. G., and Thorburn, D. (2005), ”An optimal calibration distance
leading to the optimal regression estimator,” Survey Methodology, 31(1), 95-
99.
Deville, J. C., and Särndal, C-E. (1992), ”Calibration estimators in survey
sampling,” Journal of the American Statistical Association, 87, 376-382.
Särndal, C. E., and Lundström, S. (2005), Estimation in surveys with nonresponse.
Chichester, UK: Wiley.