Miss­ing data? The TARMOS frame­work for acces­si­ble, repro­ducible analysis

Abstract

Although there is increas­ing guid­ance on how to handle miss­ing data, prac­tice is chang­ing slow­ly and misap­pre­hen­sions abound. Impor­tant­ly, the lack of trans­paren­cy around method­olog­i­cal deci­sions is threat­en­ing the valid­i­ty and repro­ducibil­i­ty of modern research.

I will describe how the recent­ly published ‘Treat­ment And Report­ing of Miss­ing data in Obser­va­tion­al Stud­ies’ (TARMOS) frame­work seeks to address these concerns, argu­ing that multi­ple impu­ta­tion (which can be used in an ever increas­ing range of settings) provides a natur­al tool to imple­ment this frame­work. The need to assist analysts with imple­ment­ing multi­ple impu­ta­tion has moti­vat­ed recent work devel­op­ing ‘mi.doc’ — an expert system in the statis­ti­cal soft­ware R, which I will briefly demonstrate. 

The ideas will be illus­trat­ed with an analy­sis of data from the UK Avon Longi­tu­di­nal Study of Parents And Chil­dren (ALSPAC), explor­ing whether there is a causal link between smok­ing at 14 years and educa­tion­al attain­ment at 16 years.

Refer­ences:

Carpen­ter, J. R., Bartlett, J. W., Morris, T. P., Wood, A. M.,Quartagno, M. and Kenward, M. G. (2023) Multi­ple Impu­ta­tion and its Appli­ca­tion (second edition). Wiley. 

Lee, K. J., Till­ing, K. M., Cornish, R. P. Little, R. J. A., Bell, M. L., Goet­ghe­beur, E., Hogan, J. W. and Carpen­ter J. R. on behalf of the STRATOS initia­tive. Frame­work for the treat­ment and report­ing of miss­ing data in obser­va­tion­al stud­ies: The Treat­ment And Report­ing of Miss­ing data in Obser­va­tion­al Stud­ies frame­work. Jour­nal of Clin­i­cal Epidemi­ol­o­gy (2021), vol 134, pages 79–88.

Speak­er

James Carpen­ter is profes­sor of medical statis­tics at the London School of Hygiene & Trop­i­cal Medi­cine and MRC Inves­ti­ga­tor in Trials Method­ol­o­gy at the MRC Clin­i­cal Trials Unit at UCL, London. His prin­ci­pal research inter­ests are coping with miss­ing data in clin­i­cal trials and complex hier­ar­chi­cal models (espe­cial­ly using multi­ple impu­ta­tion method­ol­o­gy), esti­mands and sensi­tiv­i­ty analy­sis, meta-analysis and novel trial designs.

Date

Day: Wednes­day, Septem­ber 20
Time: 10:00 am — 11:00 am
Place: AULA