Quantile estimator

Read in simulated data with different outcome-exposure relationships; y# outcome, x exposure, g instrumental variable (genotype).

use ../data/ivmwdata, clear

Plot outcome-exposure relationships

twoway line y1 x, sort(x)

twoway line y2 y3 y4 x, sort(x) legend(pos(6) rows(1))

Plot of relationships between y2, y3, y4, and x.

twoway line y5 x, sort(x)

Use with different iv commands.

ivxtile, nq(5) par(x): ivregress 2sls y2 (x = g)
qui gr export ./img/ivxtile-plot-01.svg, replace

    quantile       beta         se   medendog      lowci      uppci  
           1   .1311195   .0015519   .6070275   .1280778   .1341611  
           2   .2131875   .0009425   .9969859   .2113403   .2150348  
           3   .2827527    .001016   1.343214   .2807613   .2847441  
           4   .3815383   .0015779   1.825639   .3784456    .384631  
           5   .6742788    .054497   2.897407   .5674646   .7810929

Plot of estimated causal effect versus median of exposure in quantile.

ivxtile, nq(10) par(b1): ivtsps y2 (x = g)
qui gr export ./img/ivxtile-plot-02.svg, replace

    quantile       beta         se   medendog      lowci      uppci  
           1   .1037679   .0021337   .4384223   .0995859   .1079499  
           2   .1585745   .0015963   .7513582   .1554457   .1617032  
           3   .1981476   .0015735   .9798349   .1950635   .2012317  
           4   .2285919   .0015923   1.016022    .225471   .2317128  
           5    .263626   .0014932   1.178039   .2606993   .2665527  
           6   .3018492   .0016141   1.401061   .2986855   .3050129  
           7   .3490515   .0019169   1.642947   .3452945   .3528085  
           8   .4143449   .0022951   1.994051   .4098466   .4188433  
           9   .5128626   .0030473   2.466335   .5068899   .5188352  
          10   .8345693   .0691984   3.577371   .6989404   .9701982

Some binary outcome examples.

use ../data/ivbinoutdata, clear

ivxtile, nq(5) par(b1): ivtsps y (x = z1 z2 z3), link(logadd)
qui gr export ./img/ivxtile-plot-03.svg, replace

    quantile       beta         se    medendog      lowci      uppci  
           1   .8962642   .0908862   -.5241845   .7181272   1.074401  
           2   .5585964   .0441663    .8047578   .4720304   .6451625  
           3   .3532819   .0297168    1.851269   .2950371   .4115268  
           4   .1504481   .0224022    2.627049   .1065397   .1943565  
           5   .0427768   .0098807    4.113728   .0234107    .062143

ivxtile, nq(5) par(b1): ivtsps y (x = z1 z2 z3), link(logmult)
ivxtile, nq(5) par(b1): ivtsps y (x = z1 z2 z3), link(logit)
ivxtile, nq(5) par(b1): ivtsri y (x = z1 z2 z3), link(logadd)
ivxtile, nq(5) par(b1): ivtsri y (x = z1 z2 z3), link(logmult)
ivxtile, nq(5) par(b1): ivtsri y (x = z1 z2 z3), link(logit)
ivxtile, nq(5) par(x): ivlsmm y (x = z1 z2 z3)
ivxtile, nq(5) par(x): ivmsmm y (x = z1 z2 z3)

Save the dataset of results.

ivxtile, nq(10) par(x) saving(ivxtileres): ivreg2 y2 (x = g)