import delimited using BMI.csv, clearExamples from our IJE paper
The paper is available here (Spiller, Davies, and Palmer 2018).
mrrobust set-up
Install the mrrobust package using the user-written github package.
net install github, from("https://haghish.github.io/github/")
gitget mrrobustIf you have Stata 12 or earlier you will need to install some of these manually (see here for instructions).
Summary data description and overview
Accompanying this paper are two sets of data BMI.csv, and Height.csv, containing the set of summary estimates required for performing the BMI-serum glucose and height-serum glucose analyses respectively. Each dataset is organised into 5 columns under the following headings:
SNP: A set of identifying numbers (rsids) for each genetic variantbeta.exposure: a set of values representing the coefficient from regressing the exposure upon the genetic variant within a GWASbeta.outcome: a set of values representing the coefficient from regressing the outcome upon the genetic variant within a GWASse.exposure: a set of values representing the standard error corresponding to the coefficient in beta.exposurese.outcome: a set of values representing the standard error corresponding to the coefficient in beta.outcome.
Note Stata removes the . in the variable names when the data is imported.
In BMI.csv the exposure is standardised body mass index (BMI), and is therefore interpreted on a standard deviation scale. The summary statistics are reported by Locke et al. (2015). In Height.csv the exposure is standardised height in meters and also interpreted on a standard deviation scale. The summary statistics are reported by Wood et al. (2014).
For both analyses log transformed serum glucose was used as an outcome, reported by Shin et al. (2014). All the data was obtained from the MRBase GWAS catalogue available at https://www.mrbase.org/ (Hemani et al. 2018) which has now been superseeded by the OpenGWAS API. Genetic variants were pruned so as to be independent (\(R^2\) = 0.0001), and the effect alleles were aligned between the exposure and outcome datasets using the MRBase web application, prior to implementing mrrobust.
Stata output for each estimation method using mrrobust: BMI-Serum Glucose
Read in data
IVW
mregger betaoutcome betaexposure [aw=1/(seoutcome^2)], ivw Number of genotypes = 79
Residual standard error = 1.039
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
betaoutcome |
betaexposure | .0231866 .0079957 2.90 0.004 .0075154 .0388578
------------------------------------------------------------------------------MR-Egger
mregger betaoutcome betaexposure [aw=1/(seoutcome^2)] Number of genotypes = 79
Residual standard error = 1.046
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
betaoutcome |
slope | .0218507 .0221852 0.98 0.325 -.0216315 .0653329
_cons | .000038 .0005877 0.06 0.948 -.0011138 .0011897
------------------------------------------------------------------------------Plot of the MR-Egger model
mreggerplot betaoutcome seoutcome betaexposure seexposure
qui gr export mreggerplot-bmi.svg, width(600) replace
Weighted median
mrmedian betaoutcome seoutcome betaexposure seexposure, weighted seed(300818) Number of genotypes = 79
Replications = 1000
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
beta | .0339256 .0120248 2.82 0.005 .0103576 .0574937
------------------------------------------------------------------------------Stata output using the mode-based estimator using mrrobust: BMI-Serum Glucose
Using the mrmodalplot command, modal estimates are calculated using bandwidths of 0.25, 0.5, and 1 respectively. This command also produces three overlaid density plots for each value, as shown in the Figure.
mrmodalplot betaoutcome seoutcome betaexposure seexposure, lc(gs10 gs5 gs0) seed(300818)
qui gr export mrmodalplot-bmi.svg, width(600) replace Number of genotypes = 79
Replications = 1000
Phi = .25
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
beta | .0374507 .0424036 0.88 0.377 -.0456588 .1205602
------------------------------------------------------------------------------
Number of genotypes = 79
Replications = 1000
Phi = .5
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
beta | .0416424 .0369758 1.13 0.260 -.0308289 .1141137
------------------------------------------------------------------------------
Number of genotypes = 79
Replications = 1000
Phi = 1
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
beta | .0431816 .0281684 1.53 0.125 -.0120274 .0983906
------------------------------------------------------------------------------
Stata output for each estimation method using mrrobust: Height-Serum Glucose
Read in data
import delimited using Height.csv, clearIVW
mregger betaoutcome betaexposure [aw=1/(seoutcome^2)], ivw Number of genotypes = 367
Residual standard error = 1.044
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
betaoutcome |
betaexposure | .0015412 .0033017 0.47 0.641 -.00493 .0080124
------------------------------------------------------------------------------MR-Egger
mregger betaoutcome betaexposure [aw=1/(seoutcome^2)] Number of genotypes = 367
Residual standard error = 1.045
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
betaoutcome |
slope | -.0025878 .0091178 -0.28 0.777 -.0204584 .0152828
_cons | .0001338 .0002754 0.49 0.627 -.000406 .0006736
------------------------------------------------------------------------------Plot of the MR-Egger model
mreggerplot betaoutcome seoutcome betaexposure seexposure
qui gr export mreggerplot-height.svg, width(600) replace
Weighted median
mrmedian betaoutcome seoutcome betaexposure seexposure, weighted seed(300818) Number of genotypes = 367
Replications = 1000
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
beta | 0 .0052323 0.00 1.000 -.0102551 .0102551
------------------------------------------------------------------------------Stata output using the mode-based estimator using mrrobust: Height-Serum Glucose
mrmodalplot betaoutcome seoutcome betaexposure seexposure, lc(gs10 gs5 gs0) seed(300818)
qui gr export mrmodalplot-height.svg, width(600) replace Number of genotypes = 367
Replications = 1000
Phi = .25
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
beta | .0061368 .0245472 0.25 0.803 -.0419748 .0542484
------------------------------------------------------------------------------
Number of genotypes = 367
Replications = 1000
Phi = .5
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
beta | .0015595 .0212232 0.07 0.941 -.0400372 .0431561
------------------------------------------------------------------------------
Number of genotypes = 367
Replications = 1000
Phi = 1
------------------------------------------------------------------------------
| Coefficient Std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
beta | -.0054772 .0149074 -0.37 0.713 -.0346952 .0237408
------------------------------------------------------------------------------
References
Hemani, Gibran, Jie Zheng, Benjamin Elsworth, Kaitlin H Wade, Valeriia Haberland, Denis Baird, Charles Laurin, et al. 2018. “The MR-Base platform supports systematic causal inference across the human phenome.” eLife 7: e34408. https://doi.org/10.7554/eLife.34408.
Locke, Adam E, Bratati Kahali, Sonja I Berndt, Anne E Justice, Tune H Pers, Felix R Day, Corey Powell, et al. 2015. “Genetic Studies of Body Mass Index Yield New Insights for Obesity Biology.” Nature 518 (7538): 197–206. https://doi.org/10.1038/nature14177.
Shin, So-Youn, Eric B Fauman, Ann-Kristin Petersen, Jan Krumsiek, Rita Santos, Jie Huang, Matthias Arnold, et al. 2014. “An Atlas of Genetic Influences on Human Blood Metabolites.” Nature Genetics 46 (6): 543–50. https://doi.org/10.1038/ng.2982.
Spiller, Wes, Neil M Davies, and Tom M Palmer. 2018. “Software application profile: mrrobust—a tool for performing two-sample summary Mendelian randomization analyses.” International Journal of Epidemiology 48 (3): 684–90. https://doi.org/10.1093/ije/dyy195.
Wood, Andrew R, Tonu Esko, Jian Yang, Sailaja Vedantam, Tune H Pers, Stefan Gustafsson, Audrey Y Chu, et al. 2014. “Defining the Role of Common Variation in the Genomic and Biological Architecture of Adult Human Height.” Nature Genetics 46 (11): 1173–86. https://doi.org/10.1038/ng.3097.