1 ### EXAMPLE 1: demonstrates how measurement error can lead to a type sign error in a covariate
2 ### What kind of data invalidates fong + tyler?
3 ### Even when you have a good predictor, if it's biased against a covariate you can get the wrong sign.
4 ### Even when you include the proxy variable in the regression.
5 ### But with some ground truth and multiple imputation, you can fix it.
15 library(predictionError)
16 options(amelia.parallel="no",
20 source("simulation_base.R")
23 ### we want to estimate x -> y; x is MAR
24 ### we have x -> k; k -> w; x -> w is used to predict x via the model w.
25 ### A realistic scenario is that we have an NLP model predicting something like "racial harassment" in social media comments
26 ### The labels x are binary, but the model provides a continuous predictor
29 #### how much power do we get from the model in the first place? (sweeping N and m)
32 ## one way to do it is by adding correlation to x.obs and y that isn't in w.
33 ## in other words, the model is missing an important feature of x.obs that's related to y.
34 simulate_data <- function(N, m, B0, Bxy, Bzy, Bzx, seed, prediction_accuracy=0.73, log.likelihood.gain = 0.1){
38 # make w and y dependent
40 x <- rbinom(N, 1, plogis(Bzx*z))
42 ystar <- Bzy * z + Bxy * x + B0
43 y <- rbinom(N,1,plogis(ystar))
45 df <- data.table(x=x,y=y,ystar=ystar,z=z)
48 df <- df[sample(nrow(df), m), y.obs := y]
50 df <- df[, y.obs := y]
53 odds.y1 <- qlogis(prediction_accuracy)
54 odds.y0 <- qlogis(prediction_accuracy,lower.tail=F)
56 df[y==0,w:=plogis(rlogis(.N,odds.y0))]
57 df[y==1,w:=plogis(rlogis(.N,odds.y1))]
59 df[,w_pred := as.integer(w > 0.5)]
61 print(mean(df[x==0]$y == df[x==0]$w_pred))
62 print(mean(df[x==1]$y == df[x==1]$w_pred))
63 print(mean(df$w_pred == df$y))
67 parser <- arg_parser("Simulate data and fit corrected models")
68 parser <- add_argument(parser, "--N", default=10000, help="number of observations of w")
69 parser <- add_argument(parser, "--m", default=500, help="m the number of ground truth observations")
70 parser <- add_argument(parser, "--seed", default=17, help='seed for the rng')
71 parser <- add_argument(parser, "--outfile", help='output file', default='example_2.feather')
72 parser <- add_argument(parser, "--y_explained_variance", help='what proportion of the variance of y can be explained?', default=0.1)
73 parser <- add_argument(parser, "--prediction_accuracy", help='how accurate is the predictive model?', default=0.72)
74 ## parser <- add_argument(parser, "--x_bias_y1", help='how is the classifier biased when y = 1?', default=-0.75)
75 ## parser <- add_argument(parser, "--x_bias_y0", help='how is the classifier biased when y = 0 ?', default=0.75)
76 parser <- add_argument(parser, "--Bxy", help='coefficient of x on y', default=0.01)
77 parser <- add_argument(parser, "--Bzy", help='coeffficient of z on y', default=-0.01)
78 parser <- add_argument(parser, "--Bzx", help='coeffficient of z on x', default=-0.5)
79 parser <- add_argument(parser, "--B0", help='Base rate of y', default=0.5)
80 parser <- add_argument(parser, "--outcome_formula", help='formula for the outcome variable', default="y~x+z")
81 parser <- add_argument(parser, "--proxy_formula", help='formula for the proxy variable', default="w_pred~y")
82 parser <- add_argument(parser, "--confint_method", help='method for getting confidence intervals', default="quad")
84 args <- parse_args(parser)
92 df <- simulate_data(args$N, args$m, B0, Bxy, Bzy, Bzx, args$seed, args$prediction_accuracy)
94 # result <- list('N'=args$N,'m'=args$m,'B0'=B0,'Bxy'=Bxy,'Bzy'=Bzy, 'seed'=args$seed, 'y_explained_variance'=args$y_explained_variance, 'prediction_accuracy'=args$prediction_accuracy, 'x_bias_y0'=args$x_bias_y0,'x_bias_y1'=args$x_bias_y1,'outcome_formula' = args$outcome_formula, 'proxy_formula' = args$proxy_formula)
95 result <- list('N'=args$N,'m'=args$m,'B0'=B0,'Bxy'=Bxy,'Bzy'=Bzy, 'Bzx'=Bzx,'seed'=args$seed, 'y_explained_variance'=args$y_explained_variance, 'prediction_accuracy'=args$prediction_accuracy, 'outcome_formula' = args$outcome_formula, 'proxy_formula' = args$proxy_formula, 'confint_method'=args$confint_method)
97 outline <- run_simulation_depvar(df, result, outcome_formula = as.formula(args$outcome_formula), proxy_formula = as.formula(args$proxy_formula), confint_method=args$confint_method)
99 outfile_lock <- lock(paste0(args$outfile, '_lock'),exclusive=TRUE)
101 if(file.exists(args$outfile)){
102 logdata <- read_feather(args$outfile)
103 logdata <- rbind(logdata,as.data.table(outline),fill=TRUE)
105 logdata <- as.data.table(outline)
109 write_feather(logdata, args$outfile)