+ All Categories
Home > Documents > Assessing the accuracy of different models for combining aggregate level administrative data Dilek...

Assessing the accuracy of different models for combining aggregate level administrative data Dilek...

Date post: 20-Jan-2016
Category:
Upload: kelly-allen-caldwell
View: 213 times
Download: 0 times
Share this document with a friend
Popular Tags:
25
Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van der Heijden This research is funded by the ONS-ESRC joint studentship 1 1 1,2 1 Southampton Statistical Sciences Research Institute, University of Southampton, United Kingdom 2 Utrecht University, The Netherlands
Transcript
Page 1: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

Assessing the accuracy of different models for combining aggregate level administrative

dataDilek Yildiz

Supervisors: Peter W. F. Smith, Peter G.M. van der Heijden

This research is funded by the ONS-ESRC joint studentship

1

1 1,2

1 Southampton Statistical Sciences Research Institute, University of Southampton, United Kingdom2 Utrecht University, The Netherlands

Page 2: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

2

Outline• The Beyond 2011 Programme

• Aim

• Data sources

• Method

• Results

• Conclusion

Page 3: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

3

The Beyond 2011 Programme

• The Office for National Statistics (ONS) has been evaluating the alternative methods of collecting census data and producing small-area socio-demographic statistics.

Page 4: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

4

The Beyond 2011 Programme

The National Statistician:

“My recommendation to the Board is that the UK Statistics Authority should make the best use of all sources, combining data from an online census in 2021 and administrative data and surveys” (ONS, 2014).

Page 5: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

5

• The problems with the administrative sources:

– Collecting data from a subset of the population – Under/over coverage– People recorded with wrong age, sex or

geographic information etc.

Page 6: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

6

The Aim

• Assess the accuracy of different log-linear models for combining the aggregate level administrative data

Page 7: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

7

Data sources

m

f

Age groups

sex

LALA

18 age groups, sex and 348 local authorities (LA)

• Census Estimates ( )• All age groups

• Patient Register ( )• All age groups : Count for certain age

group, sex and LA

Page 8: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

8

• In the absence of a traditional census, instead of the census estimates it is possible to use the association structures from an alternative source such as rolling annual surveys as recommended by the ONS (2014).

Page 9: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

9

Data sources

• Census estimates are the true values

• Patient Register is biased

Patient Register*

• exceeds the census estimates by 4.3% at national level

• sex ratio (m/f) exceeds the census sex ratio for people aged between 27 and 68

• percentage difference with the census estimates are within 3% only for the 41% of local authorities

*ONS, 2012 Beyond 2011: Administrative Data Sources Report: NHS Patient Register

Percentage difference between the 2011 Patient Register and the 2011 Census estimates for total population

Page 10: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

10

Recent papers• Raymer and Rogers (2007) - Using age and spatial

flow structures in the indirect estimation of migration streams.

• Raymer, et al. (2007) - Combining census and registration data to estimate detailed elderly migration flows in England and Wales.

• Raymer, et al. (2009) - Combining census and registration data to analyse ethnic migration patterns in England from 1991 to 2007.

• Smith, et al. (2010) - Combining available migration data in England to study economic activity flows over time.

Page 11: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

11

Method

• (Total) model:

• (AS) model:

• (AS,L) model:

• (AS,SL) model:

• (AS,AL) model:

Page 12: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

12

Mean percentage differences for age groups, total population

0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+

-6.0

-4.0

-2.0

0.0

2.0

4.0

6.0

8.0

10.0

12.0

PR Total model AS modelAS,L model AS,SL model AS, AL model

Page 13: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

13

Mean percentage differences for age groups, males

0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+

-6.0

-4.0

-2.0

0.0

2.0

4.0

6.0

8.0

10.0

12.0

PR Total model AS modelAS,L model AS,SL model AS, AL model

Page 14: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

14

Mean percentage differences for age groups, females

0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+

-6.0

-4.0

-2.0

0.0

2.0

4.0

6.0

8.0

10.0

12.0

PR Total model AS modelAS,L model AS,SL model AS, AL model

Page 15: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

15

Total populationThe Patient Register AS model

Page 16: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

16

20-24 Males The Patient Register AS,AL model

Page 17: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

17

40-44 Males

The Patient Register AS,AL model

Page 18: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

18

70-74 MalesPatient Register AS,AL model

Page 19: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

19

20-24 Females The Patient Register AS,AL model

Page 20: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

20

40-44 FemalesThe Patient Register AS,SL model

Page 21: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

21

70-74 FemalesThe Patient Register AS,AL model

Page 22: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

22

Percentage of local authorities within 3.8% of Census Estimates

PR Total AS AS,L AS,SL AS,AL

Total population 57 88 91 100 100 100

20-24 Males 23 32 39 28 28 76

35-39 Males 16 34 39 52 58 87

40-44 Males 12 42 43 59 67 85

70-74 Males 67 45 82 79 74 97

20-24 Females 24 42 52 49 52 76

35-39 Females 57 66 66 90 87 86

40-44 Females 72 64 86 91 95 86

70-74 Females 78 34 83 82 89 98

Page 23: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

23

Conclusion

• The estimates get better when more corrections are made but complicated models require more association information about the population.

• It is possible to obtain association structures from another source in the future such as 4% rolling annual surveys as proposed by the ONS.

• This research can also be extended to use different age groups (such as 0-19, 20-39, 40-59, 60+).

Page 24: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

24

References• Office for National Statistics (ONS) (2014), 27 March 2014 - The census and future

provision of population statistics in England and Wales: Recommendation from the National Statistician and Chief Executive of the UK Statistics Authority, Office for National Statistics.

• Office for National Statistics (ONS) (2012), Beyond 2011: Administrative Data Sources Report: NHS Patient Register, Office for National Statistics.

• Office for National Statistics (ONS) (2012), The 2011 Census coverage assessment and adjustment process, Office for National Statistics.

• Raymer, J. and Rogers, A. (2007) Using age and spatial flow structures in the indirect estimation of migration streams. Demography. 44, 199-223. DOI: 10.1353/dem.2007.0016.

• Raymer, J., Abel, G. and Smith, P.W. F. (2007), Combining census and registration data to estimate detailed elderly migration flows in England and Wales, Journal of the Royal Statistical Society, Series A, 170(4), 891-908.

• Raymer, J., Smith, P.W. F., and Guilietti, C. (2009), Combining census and registration data to analyse ethnic migration patterns in England from 1991 to 2007, Population, Space and Place, 17, 73-88.

• Smith, P.W. F., Raymer, J., and Guilietti, C. (2010) Combining available migration data in England to study economic activity flows over time. Journal of the Royal Statistical Society, Series A (Statistics in Society). 173(4), 733-753. DOI: 10.1111/j.1467-985X.2009.00630.x.

Page 25: Assessing the accuracy of different models for combining aggregate level administrative data Dilek Yildiz Supervisors: Peter W. F. Smith, Peter G.M. van.

25

Thank you for your attention


Recommended