Limits...
A toolkit for measurement error correction, with a focus on nutritional epidemiology.

Keogh RH, White IR - Stat Med (2014)

Bottom Line: To correct for the effects of measurement error, information additional to the main study data is required.We emphasize practical application of the methods in nutritional epidemiology and other fields.The methods are illustrated using the data from a study of the association between fibre intake and colorectal cancer, where fibre intake is measured using a diet diary and repeated measures are available for a subset.

View Article: PubMed Central - PubMed

Affiliation: Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, U.K.

Show MeSH

Related in: MedlinePlus

Results from a case–control study within the EPIC-Norfolk cohort. Log odds ratio estimates within quintiles of exposure: (i) naive method: naive log odds ratios within quintiles of observed log scale fibre intake plotted against mean observed exposure within quintile, (ii) MacMahon's method: naive log odds ratios are plotted against estimated usual intake within observed quintiles and (iii) moment reconstruction: log odds ratios within quintiles of true exposure (estimated using moment reconstruction) are plotted against mean estimated true exposure with those quintiles. Bars give 95% confidence intervals relative to the lowest quintile.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4285313&req=5

fig02: Results from a case–control study within the EPIC-Norfolk cohort. Log odds ratio estimates within quintiles of exposure: (i) naive method: naive log odds ratios within quintiles of observed log scale fibre intake plotted against mean observed exposure within quintile, (ii) MacMahon's method: naive log odds ratios are plotted against estimated usual intake within observed quintiles and (iii) moment reconstruction: log odds ratios within quintiles of true exposure (estimated using moment reconstruction) are plotted against mean estimated true exposure with those quintiles. Bars give 95% confidence intervals relative to the lowest quintile.

Mentions: We also applied the correction methods for categorized exposures, described in Section 0018. Figure 2 shows the results from a naive analysis using quintiles of exposure, from MacMahon's method, and by imputing the underlying continuous exposure (on the log scale here) and categorizing that for use in the exposure-outcome analysis. In the uncorrected plot, there is some suggestion of a non-linear association. The results using the correction methods suggest a more non-linear relationship compared with the naive analysis, which is what we would expect 50.


A toolkit for measurement error correction, with a focus on nutritional epidemiology.

Keogh RH, White IR - Stat Med (2014)

Results from a case–control study within the EPIC-Norfolk cohort. Log odds ratio estimates within quintiles of exposure: (i) naive method: naive log odds ratios within quintiles of observed log scale fibre intake plotted against mean observed exposure within quintile, (ii) MacMahon's method: naive log odds ratios are plotted against estimated usual intake within observed quintiles and (iii) moment reconstruction: log odds ratios within quintiles of true exposure (estimated using moment reconstruction) are plotted against mean estimated true exposure with those quintiles. Bars give 95% confidence intervals relative to the lowest quintile.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4285313&req=5

fig02: Results from a case–control study within the EPIC-Norfolk cohort. Log odds ratio estimates within quintiles of exposure: (i) naive method: naive log odds ratios within quintiles of observed log scale fibre intake plotted against mean observed exposure within quintile, (ii) MacMahon's method: naive log odds ratios are plotted against estimated usual intake within observed quintiles and (iii) moment reconstruction: log odds ratios within quintiles of true exposure (estimated using moment reconstruction) are plotted against mean estimated true exposure with those quintiles. Bars give 95% confidence intervals relative to the lowest quintile.
Mentions: We also applied the correction methods for categorized exposures, described in Section 0018. Figure 2 shows the results from a naive analysis using quintiles of exposure, from MacMahon's method, and by imputing the underlying continuous exposure (on the log scale here) and categorizing that for use in the exposure-outcome analysis. In the uncorrected plot, there is some suggestion of a non-linear association. The results using the correction methods suggest a more non-linear relationship compared with the naive analysis, which is what we would expect 50.

Bottom Line: To correct for the effects of measurement error, information additional to the main study data is required.We emphasize practical application of the methods in nutritional epidemiology and other fields.The methods are illustrated using the data from a study of the association between fibre intake and colorectal cancer, where fibre intake is measured using a diet diary and repeated measures are available for a subset.

View Article: PubMed Central - PubMed

Affiliation: Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, U.K.

Show MeSH
Related in: MedlinePlus