I am using the following code to generate data, and i am estimating regression models across a list of variables (covar1 and covar2). I have also created confidence intervals for the coefficients and merged them together.
I have been examining all sorts of examples here and on other sites, but i can't seem to accomplish what i want. I want to stack the results for each covar into a single data frame, labeling each cluster of results by the covar it is attributable to (i.e., "covar1" and "covar2"). Here is the code for generating data and results using lapply:
##creating a fake dataset (N=1000, 500 at treated, 500 at control group)
#outcome variable
outcome <- c(rnorm(500, mean = 50, sd = 10), rnorm(500, mean = 70, sd = 10))
#running variable
running.var <- seq(0, 1, by = .0001)
running.var <- sample(running.var, size = 1000, replace = T)
##Put negative values for the running variable in the control group
running.var[1:500] <- -running.var[1:500]
#treatment indicator (just a binary variable indicating treated and control groups)
treat.ind <- c(rep(0,500), rep(1,500))
#create covariates
set.seed(123)
covar1 <- c(rnorm(500, mean = 50, sd = 10), rnorm(500, mean = 50, sd = 20))
covar2 <- c(rnorm(500, mean = 10, sd = 20), rnorm(500, mean = 10, sd = 30))
data <- data.frame(cbind(outcome, running.var, treat.ind, covar1, covar2))
data$treat.ind <- as.factor(data$treat.ind)
#Bundle the covariates names together
covars <- c("covar1", "covar2")
#loop over them using a convenient feature of the "as.formula" function
models <- lapply(covars, function(x){
regres <- lm(as.formula(paste(x," ~ running.var + treat.ind",sep = "")), data = d)
ci <-confint(regres, level=0.95)
regres_ci <- cbind(summary(regres)$coefficient, ci)
})
names(models) <- covars
print(models)
Any nudge in the right direction, or link to a post i just haven't come across, is greatly appreciated.
You can use do.call were de second argument is a list (like in here):
do.call(rbind, models)
I made a (possible) improve to your lapply function. This way you can save the estimated parameters and the variables in a data.frame:
models <- lapply(covars, function(x){
regres <- lm(as.formula(paste(x," ~ running.var + treat.ind",sep = "")), data = data)
ci <-confint(regres, level=0.95)
regres_ci <- data.frame(covar=x,param=rownames(summary(regres)$coefficient),
summary(regres)$coefficient, ci)
})
do.call(rbind,models)
Related
The code I have so far is written below. This is simulation so there is no actual data. I have two questions:
I have two vectors (treat and cont) but I need to put them into one single vector which I did (vect), however, I need another vector that is coding for treatment vs. control. How do I do that?
For my model (model) I need to fit a linear model testing for a treatment effect but I don't know how to add that effect into what I have or is that what it is testing in the code I have?
library(car)
treat=rnorm(3, mean = 460, sd = 110)
treat
cont=rnorm(3, mean = 415, sd = 110)
cont
vect=c(treat, cont)
vect
nsims = 1000
p.value.saved = coeff.saved = vector()
for (i in 1:nsims) {
treat=rnorm(3, mean = 460, sd = 110)
cont=rnorm(3, mean = 415, sd = 110)
vect=c(treat, cont)
model = glm(treat ~ cont, family = poisson)
p.value.saved[i] = Anova(model)$P[1]
coeff.saved[i] = coef(model)
}
Thank you!
Something like this? (note that you'll get a bunch of warnings for running a poisson regression against continuous data.
n <- 3
nsims <- 10
do.call(
rbind,
lapply(1:nsims, function(.) {
treat <- rnorm(n, mean = 460, sd = 110)
cont <- rnorm(n, mean = 415, sd = 110)
# Instead of vect
df <- data.frame(
y = c(treat, cont),
x = rep(c("treat", "cont"), each = n)
)
# Model the values vs treatment indicator
model <- glm(y ~ x, data = df, family = poisson)
# Extract the model's p-value and coefficient of treatment.
data.frame(p = car::Anova(model)$P, coef = coef(model)[2])
})
)
The first creates the string and the second bit will combine them. In your example they are both length 3, hence the 3 repetition in rep("trt",3)
treat_lab = c(rep("control", 3),rep("trt", 3))
treatment <- cbind(treat_lab,c(treat,cont))
I need to plot the statistical power vs. the number of replicates and in this case the number of replicates (n) is 3, but I can't figure out how to plot it.
This is what I have:
library(car)
n <- 3
nsims <- 1000
p = coef = vector()
for (i in 1:nsims) {
treat <- rnorm(n, mean = 460, sd = 110)
cont <- rnorm(n, mean = 415, sd = 110)
df <- data.frame(
y = c(treat, cont),
x = rep(c("treat", "cont"), each = n)
)
model <- glm(y ~ x, data = df)
p[i] = Anova(model)$P
coef[i] = coef(model)[2]
}
hist(p, col = 'skyblue')
sum(p < 0.05)/nsims
Can someone help me plot this?
Also, I need to calculate the mean of the coefficients using only models where p < 0.05. This is simulating the following process: if you perform the experiment, and p > 0.05, you report 'no effect’, but if p < 0.05 you report ‘significant effect’. But I'm not sure how to set that up from what I have.
Would I just do this?
mean(coef)
But I don't know how to include only those with p < 0.05.
Thank you!
Disclaimer: I spend a decent amount of time simulating experiments for work so I have strong opinions on this.
If that's everything because it's for a study assignment then fine, if you are planning to go further with this I recommend
adding the tidyverse to your arsenal.
Encapsulating functionality
First allows me to put a single iteration into a function to decouple its logic from the result subsetting (the encapsulation).
sim <- function(n) {
treat <- rnorm(n, 460, 110)
cont <- rnorm(n, 415, 110)
data <- data.frame(y = c(treat, cont), x = rep(c("treat", "cont"), each = n))
model <- glm(y ~ x, data = data)
p <- car::Anova(model)$P
coef <- coef(model)[2]
data.frame(n, p, coef)
}
Now we can simulate
nsims <- 1000
sims <- do.call(
rbind,
# We are now using the parameter as opposed to the previous post.
lapply(
rep(c(3, 5, 10, 20, 50, 100), each = nsims),
sim
)
)
# Aggregations
power_smry <- aggregate(p ~ n, sims, function(x) {mean(x < 0.05)})
coef_smry <- aggregate(coef ~ n, sims[sims$p < 0.05, ], mean)
# Plots
plot(p ~ n, data = power_smry
If you do this in the tidyverse this is one possible approach
crossing(
n = rep(c(3, 5, 10, 20, 50, 100))
# Add any number of other inputs here that you want to explore (like lift).
) %>%
rowwise() %>%
# This looks complicated but will be less so if you have multiple
# varying hyperparameters defined in crossing.
mutate(results = list(bind_rows(rerun(nsims, sim(n))))) %>%
pull(results) %>%
bind_rows() %>%
group_by(n) %>%
# The more metrics you want to summarize in different ways the easier compared to base.
summarize(
power = mean(p < 0.05),
coef = mean(coef[p < 0.05])
)
I have a data set (df) in this format
index <- runif(n = 100,min = 0, max = 1)
type1 <- rep("low", 50)
type2 <- rep("high", 50)
type <- c(type1,type2)
level1 <- rep("single", 25)
level2 <- rep("multiple", 25)
level3 <- rep("single", 25)
level4 <- rep("multiple", 25)
level <- c(level1,level2,level3,level4)
block <- rep(1:5, 10)
set <- rep(1:5, 10)
df <- data.frame("index" = index,"type" = type, "level" = level, "block" = block, "set" = set)
df$block <- as.factor(df$block)
df$set <- as.factor(df$set)
I want to create a model that looks like like this
model <- lmer(index ~ type * level + (1|block) + (1|set), data = df)
However, in my original data the fit is bad because the data is bound between 0 and 1. I want to bootstrap this mixed effects model. Any idea on how to achieve boot-strapping for such a model? I want to compare this this full model with sub-models eg. without interaction, or with level or type alone. I also want with confidence intervals for the final model
The confint() function has a method for merMod objects. The following should work:
confint(model, method = "boot", nsim = 1000)
And with multiple CPUs:
confint(model, method = "boot", nsim = 1000,
parallel = "multicore", ncpus = 8)
I am having trouble passing data to forecast.lm in a dplyr do. I want to make several models based on a a factor - hour - and the forecaste these models using new data.
Building on previous excellent examples here is my data example:
require(dplyr)
require(forecast)
# Training set
df.h <- data.frame(
hour = factor(rep(1:24, each = 100)),
price = runif(2400, min = -10, max = 125),
wind = runif(2400, min = 0, max = 2500),
temp = runif(2400, min = - 10, max = 25)
)
# Forecasting set
df.f <- data.frame(
hour = factor(rep(1:24, each = 10)),
wind = runif(240, min = 0, max = 2500),
temp = runif(240, min = - 10, max = 25)
)
# Bind training & forecasting
df <- rbind(df.h, data.frame(df.f, price=NA))
# Do a training model and then forecast using the new data
df <- rbind(df.h, data.frame(df.f, price=NA))
res <- group_by(df, hour) %>% do({
hist <- .[!is.na(.$price), ]
fore <- .[is.na(.$price), c('hour', 'wind', 'temp')]
fit <- Arima(hist$price, xreg = hist[,3:4], order = c(1,1,0))
data.frame(fore[], price=forecast.Arima(fit, xreg = fore[ ,2:3])$mean)
})
res
This works excellently with a time series model, but using a lm I have problem passing the data into the forecasting part.
My corresponding lm example looks like this:
res <- group_by(df, hour) %>% do({
hist <- .[!is.na(.$price), ]
fore <- .[is.na(.$price), c('hour', 'wind', 'temp')]
fit <- lm(hist$price ~ wind + temp, data = hist)
data.frame(fore[], price = forecast.lm(fit, newdata = fore[, 2:3])$mean)
})
The problem is that I cant' get data into the newdata = function. If you add hist$ in the fit section, you can't reference the forecast data, and for some reason if you add data = fore it can't find it - but it can in the time series example.
The problem is that forecast.lm expects that fit has a data component. If you use glm or tslm, that is true. But lm objects don't generally have a data component. So you need to manually add fit$data <- hist for forecast.lm to work properly.
res <- group_by(df, hour) %>% do({
hist <- .[!is.na(.$price), ]
fore <- .[is.na(.$price), c('hour', 'wind', 'temp')]
fit <- lm(price ~ wind + temp, data = hist)
fit$data <- hist # have to add data manually
data.frame(fore[], price = forecast.lm(fit, newdata = fore[, 2:3])$mean)
})
This is actually a known issue.
I have a data frame and I'm trying to create a new variable in the data frame that has the quantiles of a continuous variable var1, for each level of a factor strata.
# some data
set.seed(472)
dat <- data.frame(var1 = rnorm(50, 10, 3)^2,
strata = factor(sample(LETTERS[1:5], size = 50, replace = TRUE))
)
# function to get quantiles
qfun <- function(x, q = 5) {
quantile <- cut(x, breaks = quantile(x, probs = 0:q/q),
include.lowest = TRUE, labels = 1:q)
quantile
}
I tried using two methods, neither of which produce a usable result. Firstly, I tried using aggregate to apply qfun to each level of strata:
qdat <- with(dat, aggregate(var1, list(strata), FUN = qfun))
This returns the quantiles by factor level, but the output is hard to coerce back into a data frame (e.g., using unlist does not line the new variable values up with the correct rows in the data frame).
A second approach was to do this in steps:
tmp1 <- with(dat, split(var1, strata))
tmp2 <- lapply(tmp1, qfun)
tmp3 <- unlist(tmp2)
dat$quintiles <- tmp3
Again, this calculates the quantiles correctly for each factor level, but obviously, as with aggregate they aren't in the correct order in the data frame. We can check this by putting the quantile "bins" into the data frame.
# get quantile bins
qfun2 <- function(x, q = 5) {
quantile <- cut(x, breaks = quantile(x, probs = 0:q/q),
include.lowest = TRUE)
quantile
}
tmp11 <- with(dat, split(var1, strata))
tmp22 <- lapply(tmp11, qfun2)
tmp33 <- unlist(tmp22)
dat$quintiles2 <- tmp33
Many of the values of var1 are outside of the bins of quantile2. I feel like i'm missing something simple. Any suggestions would be greatly appreciated.
I think your issue is that you don't really want to aggregate, but use ave, (or data.table or plyr)
qdat <- transform(dat, qq = ave(var1, strata, FUN = qfun))
#using plyr
library(plyr)
qdat <- ddply(dat, .(strata), mutate, qq = qfun(var1))
#using data.table (my preference)
dat[, qq := qfun(var1), by = strata]
Aggregate usually implies returning an object that is smaller that the original. (inthis case you were getting a data.frame where x was a list of 1 element for each strata.
Use ave on your dat data frame. Full example with your simulated data and qfun function:
# some data
set.seed(472)
dat <- data.frame(var1 = rnorm(50, 10, 3)^2,
strata = factor(sample(LETTERS[1:5], size = 50, replace = TRUE))
)
# function to get quantiles
qfun <- function(x, q = 5) {
quantile <- cut(x, breaks = quantile(x, probs = 0:q/q),
include.lowest = TRUE, labels = 1:q)
quantile
}
And my addition...
dat$q <- ave(dat$var1,dat$strata,FUN=qfun)