8 3/20 Lab VII | Lists, Loops, & Functions in R
8.1 Preparation
## Packages
library(plyr)
library(tidyverse)
library(readxl)
library(readr)
library(lubridate)
library(knitr)Differences in policies and personnel associated with presidential candidates could affect the profitability of firms in different ways, with stocks potentially rising and falling with the expected outlooks for political candidates. In this report we analyze stocks that have been identified as potentially responsive to the political fortunes of Donald Trump and Joe Biden.
8.1.1 Data
Our stock market data consists of daily closing prices of 31 stocks that were identified by Kiplinger as being politically exposed. These stocks cover a broad range of sectors potentially influenced by politics, including tech stocks, energy (both oil and renewable oriented firms), health care, marijuana, gold and foreign-oriented ETFs.
Our measure of political expectations comes from the PredictIt betting market in which investors buy shares in (among other things) presidential candidates. Share are worth $1 if the candidate wins the general election. The price for a candidate on a given day is taken to be the market’s estimate of the probability of victory for that candidate.
8.3 Create data frame with daily price data for Trump in 2020
## Creating daily price data for Trump
trump2020 <- pres_mkt %>%
filter(year(date) > 2019 & name == "Donald Trump") %>%
dplyr::select(c(date, "Trump" = price))8.4 Merge the stock price and Trump price data
## Joining the two datasets
stk_pres <- left_join(stk, trump2020, by = "date")8.5 Create a list of stock ticker names (do not include DJI as a ticker)
## Creating list of ticker names
stk_tickers <-
c(names(stk)[!names(stk) %in% c("date", "DJI")])
## Creating number of stocks
stk_num <- length(stk_tickers)8.6 Create a function that calculates daily percent change. Use the lag function and the mutate_at function. Make sure to check that your function worked.
## Price Change Function
pct_change <- function(x){
(x - lag(x))/lag(x)
}
## Daily Change Dataset
daily_df <- stk_pres %>%
mutate_at(c("DJI", "Trump", stk_tickers), pct_change) %>%
dplyr::select(c(date, c("DJI", "Trump", stk_tickers)))8.7 Loop thru list of tickers and run regressions in which daily change in stock price is a function of change in DJIA and change in Trump price. Create a data frame that stores the coefficient, standard error and t-stat for the Trump variable for each stock. Include a column that has the stock ticker in that data frame as well.
## Preparing Data Frame to store results
ols_results <- data.frame("row" = 1:stk_num,
"ticker" = NA,
"coef" = NA,
"se" = NA,
"tStat" = NA)
## Looping Though Regressions
for(i in 1:stk_num){
daily_df$temp <- unlist(daily_df[, i + 3])
ols.1 <- lm(temp ~ DJI + Trump, data = daily_df)
ols_results[i, "ticker"] <- stk_tickers[i]
ols_results[i, 3:5] <- round(summary(ols.1)$coefficients["Trump", 1:3], 3)
}
## Showing results
head(ols_results)## row ticker coef se tStat
## 1 1 AAPL -0.015 0.011 -1.304
## 2 2 NFLX -0.036 0.016 -2.208
## 3 3 AMZN -0.034 0.013 -2.643
## 4 4 K -0.003 0.011 -0.259
## 5 5 TSLA -0.004 0.031 -0.115
## 6 6 LMT 0.000 0.011 -0.0268.8 The questions below are not required for full credit but I encourage you to try!
8.8.1 Use list apply - lapply() - to regress stock price on Trump.
## Now with lapply()
ols_results2 <- lapply(stk_tickers, function(x){
daily_df$temp = unlist(daily_df[, which(names(daily_df) == x)])
ols.1 <- lm(temp ~ DJI + Trump, data = daily_df)
})
## Showing Results
head(ols_results2)## [[1]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001889 0.980104 -0.014758
##
##
## [[2]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001903 0.535397 -0.035819
##
##
## [[3]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.002097 0.532408 -0.033624
##
##
## [[4]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.0003296 0.4133016 -0.0028461
##
##
## [[5]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.009543 1.042136 -0.003596
##
##
## [[6]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.0002786 0.8337098 -0.00028568.8.2 Use tapply to find the average price of each presidential candidate over the course of the campaign.
## With tapply()
tapply(pres_mkt$price, pres_mkt$name, mean)## Amy Klobuchar Andrew Cuomo Andrew Yang Bernie Sanders Beto O'Rourke Cory Booker
## 0.02166512 0.01606446 0.03927786 0.09528414 0.03062162 0.03407125
## Donald Trump Elizabeth Warren Hillary Clinton Howie Hawkins Jo Jorgensen Joe Biden
## 0.38840543 0.07018660 0.02375566 0.01000000 0.01000000 0.24708227
## John Kasich Kamala Harris Kanye West Kirsten Gillibrand Mark Cuban Mark Zuckerberg
## 0.01776930 0.06757422 0.01000000 0.03126378 0.01106383 0.01024052
## Michael Bloomberg Mike Pence Nikki Haley Paul Ryan Pete Buttigieg Sherrod Brown
## 0.02249647 0.03799830 0.01691027 0.01322849 0.04516535 0.01289140
## Tom Steyer Tulsi Gabbard
## 0.01000000 0.012211068.8.3 Now use map() in library(purrr) to regress stock price on Trump.
## The function from above
func_stock <- function(x){
daily_df$temp = unlist(daily_df[, which(names(daily_df) == x)])
ols.1 <- lm(temp ~ DJI + Trump, data = daily_df)
}
## Now with map()
map(.x=stk_tickers, .f=func_stock)## [[1]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001889 0.980104 -0.014758
##
##
## [[2]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001903 0.535397 -0.035819
##
##
## [[3]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.002097 0.532408 -0.033624
##
##
## [[4]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.0003296 0.4133016 -0.0028461
##
##
## [[5]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.009543 1.042136 -0.003596
##
##
## [[6]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.0002786 0.8337098 -0.0002856
##
##
## [[7]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.0003462 0.9118574 -0.0070218
##
##
## [[8]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.0007476 1.3508059 0.0489526
##
##
## [[9]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001328 0.658680 -0.008615
##
##
## [[10]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.002651 1.056977 0.028651
##
##
## [[11]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001382 1.011709 -0.010030
##
##
## [[12]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.00125 0.86138 -0.04332
##
##
## [[13]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.0008178 0.8435450 -0.0202986
##
##
## [[14]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.003142 1.067967 0.019232
##
##
## [[15]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001017 0.463254 -0.002714
##
##
## [[16]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.002939 0.866421 0.005825
##
##
## [[17]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.0001886 1.0378297 0.0003729
##
##
## [[18]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 4.747e-05 1.067e+00 2.402e-02
##
##
## [[19]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.0005725 0.9488423 0.0269447
##
##
## [[20]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.0005572 0.5019176 -0.0012461
##
##
## [[21]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001302 1.009206 0.004588
##
##
## [[22]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.0006483 0.8554046 0.0061128
##
##
## [[23]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001798 0.893112 -0.010831
##
##
## [[24]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.001051 0.810440 -0.008493
##
##
## [[25]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.002559 0.937766 0.005098
##
##
## [[26]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.01192 0.69831 0.05660
##
##
## [[27]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.0008981 1.0917988 -0.0242325
##
##
## [[28]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.001052 1.176215 0.019945
##
##
## [[29]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## -0.002581 0.959133 0.031636
##
##
## [[30]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.0001306 0.8001097 -0.0106032
##
##
## [[31]]
##
## Call:
## lm(formula = temp ~ DJI + Trump, data = daily_df)
##
## Coefficients:
## (Intercept) DJI Trump
## 0.0008454 0.0707473 -0.01240078.8.4 Create your own simple regression function. To avoid issues of NAs, simulate some data and see if it worked!
## Creating constituent functions
## Beta Hat Function
beta_hat <- function(x,y) {
(sum((x-mean(x))*(y-mean(y))))/
(sum((x-mean(x))^2))
}
## Intercept Function
constant <- function(x,y,b){
mean(y) - b*mean(x)
}
## Fitted Values Function
fit_val <- function(x, y, b) {
constant(x=x,y=y,b=b) + beta_hat(x,y) * x
}
## R squared Function
r2 <- function(fitted, y) {
sum((fitted - mean(y))^2)/
sum((y-mean(y))^2)
}
## Standard Error Function
se <- function(fitted, x, y) {
sqrt(sum((fitted-y)^2)/length(y)/
sum((x-mean(x))^2))
}
## T Stat Function
t <- function(beta, null=0, se) {
(beta-null)/
se
}
## P-Value Function
pval <- function(t,df) {
pt(q=t, df, lower.tail = F)*2
}
## OLS Function
ols <- function(X,Y) {
b <- beta_hat(X, Y)
constant <- constant(x=X, y=Y, b)
fitted <- fit_val(x=X,y=Y,b=b)
r_squared <- r2(fitted, y=Y)
stand_err <- se(fitted=fitted, x=X, y=Y)
t_stat <- t(beta=b, se=stand_err)
p <- pval(t=t_stat, df=(length(X)-1))
tibble(b, stand_err, t_stat, p)
}
## Creating Data
x <- rnorm(1000, 15, 2)
y <- x*17 + rnorm(1000, 15, 12)
## Testing
ols(x, y)## # A tibble: 1 × 4
## b stand_err t_stat p
## <dbl> <dbl> <dbl> <dbl>
## 1 17.2 0.186 92.2 0summary(lm(y~x))##
## Call:
## lm(formula = y ~ x)
##
## Residuals:
## Min 1Q Median 3Q Max
## -37.944 -7.582 -0.012 7.890 44.258
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 13.3346 2.8372 4.70 2.97e-06 ***
## x 17.1512 0.1862 92.13 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 11.52 on 998 degrees of freedom
## Multiple R-squared: 0.8948, Adjusted R-squared: 0.8947
## F-statistic: 8489 on 1 and 998 DF, p-value: < 2.2e-16