hermiter

1 What does hermiter do?

hermiter facilitates the estimation of the probability density function and cumulative distribution function in univariate and bivariate settings using Hermite series based estimators. In addition, hermiter allows the estimation of the quantile function in the univariate case and nonparametric correlation coefficients in the bivariate case. The package is applicable to streaming, batch and grouped data. The core methods of the package are written in C++ for speed.

These estimators are particularly useful in the sequential setting (both stationary and non-stationary data streams). In addition, they are useful in efficient, one-pass batch estimation which is particularly relevant in the context of large data sets. Finally, the Hermite series based estimators are applicable in decentralized (distributed) settings in that estimators formed on subsets of the data can be consistently merged. The Hermite series based estimators have the distinct advantage of being able to estimate the full density function, distribution function and quantile function (univariate setting) along with the Spearman Rho and Kendall Tau correlation coefficients (bivariate setting) in an online manner. The theoretical and empirical properties of most of these estimators have been studied in-depth in the articles below. The investigations demonstrate that the Hermite series based estimators are particularly effective in distribution function, quantile function and Spearman correlation estimation.

A summary of the estimators and algorithms in hermiter can be found in the article below.

2 Load Packages

In order to utilize the hermiter package, the package must be loaded using the following command:

library(hermiter)

Other packages that are used in this vignette are loaded as follows:

library(magrittr)
library(ggplot2)
library(dplyr)
library(data.table)
library(DT)
library(mvtnorm)
library(patchwork)
library(colorspace)

3 Construct Estimator

A hermite_estimator S3 object is constructed as below. The argument, N, adjusts the number of terms in the Hermite series based estimator and controls the trade-off between bias and variance. A lower N value implies a higher bias but lower variance and vice versa for higher values of N. The argument, standardize, controls whether or not to standardize observations before applying the estimator. Standardization usually yields better results and is recommended for most estimation settings.

A univariate estimator is constructed as follows (note that the default estimator type is univariate, so this argument does not need to be explicitly set):

hermite_est <- hermite_estimator(N=10, standardize=TRUE, 
                                 est_type = "univariate")

Similarly for constructing a bivariate estimator:

hermite_est <- hermite_estimator(N=10, standardize=TRUE, 
                                 est_type = "bivariate")

4 Batch Estimator Updating

A hermite_estimator object can be initialized with a batch of observations as below.

4.1 Standard syntax

For univariate observations:

observations <- rlogis(n=1000)
hermite_est <- hermite_estimator(N=10, standardize=TRUE, observations = 
                                   observations)

For bivariate observations:

observations <- matrix(data = rnorm(2000),nrow = 1000, ncol=2)
hermite_est <- hermite_estimator(N=10, standardize=TRUE, 
                                 est_type = "bivariate", observations = 
                                   observations)

Functional piped syntax can also be used as below provided the magrittr package is installed.

5 Sequential Estimator Updating

In the sequential setting, observations are revealed one at a time. A hermite_estimator object can be updated sequentially with a single new observation by utilizing the update_sequential method. Note that when updating the Hermite series based estimator sequentially, observations are also standardized sequentially if the standardize argument is set to true in the constructor.

5.1 Standard syntax

For univariate observations:

observations <- rlogis(n=1000)
hermite_est <- hermite_estimator(N=10, standardize=TRUE)
hermite_est <- update_sequential(hermite_est,observations)

For bivariate observations:

observations <- matrix(data = rnorm(2000),nrow = 1000, ncol=2)
hermite_est <- hermite_estimator(N=10, standardize=TRUE, 
                                 est_type = "bivariate")
hermite_est <- update_sequential(hermite_est,observations)

5.2 Piped syntax

For univariate observations:

observations <- rlogis(n=1000)
hermite_est <- hermite_estimator(N=10, standardize=TRUE)
hermite_est <- hermite_est %>% update_sequential(observations)

For bivariate observations:

observations <- matrix(data = rnorm(2000),nrow = 1000, ncol=2)
hermite_est <- hermite_estimator(N=10, standardize=TRUE, 
                                 est_type = "bivariate")
hermite_est <- hermite_est %>% update_sequential(observations)

6 Merging Hermite Estimators

Hermite series based estimators can be consistently combined/merged in both the univariate and bivariate settings. In particular, when standardize = FALSE, the results obtained from combining/merging distinct hermite_estimators updated on subsets of a data set are exactly equal to those obtained by constructing a single hermite_estimator and updating on the full data set (corresponding to the concatenation of the aforementioned subsets). This holds true for the pdf, cdf and quantile results in the univariate case and the pdf, cdf and nonparametric correlation results in the bivariate case. When standardize = TRUE, the equivalence is no longer exact, but is accurate enough to be practically useful. Combining/merging hermite_estimators is illustrated below.

For the univariate case:

observations_1 <- rlogis(n=1000)
observations_2 <- rlogis(n=1000)
hermite_est_1 <- hermite_estimator(N=10, standardize=TRUE, 
                                   observations = observations_1)
hermite_est_2 <- hermite_estimator(N=10, standardize=TRUE, 
                                   observations = observations_2)
hermite_est_merged <- merge_hermite(list(hermite_est_1,hermite_est_2))

For the bivariate case:

observations_1 <- matrix(data = rnorm(2000),nrow = 1000, ncol=2)
observations_2 <- matrix(data = rnorm(2000),nrow = 1000, ncol=2)
hermite_est_1 <- hermite_estimator(N=10, standardize=TRUE, 
                                 est_type = "bivariate", 
                                 observations = observations_1)
hermite_est_2 <- hermite_estimator(N=10, standardize=TRUE, 
                                 est_type = "bivariate", 
                                 observations = observations_2)
hermite_est_merged <- merge_hermite(list(hermite_est_1,hermite_est_2))

The ability to combine/merge estimators is particularly useful in applications involving grouped data (as we will illustrate below).

7 Estimate univariate pdf, cdf and quantile function

The central advantage of Hermite series based estimators is that they can be updated in a sequential/one-pass manner as above and subsequently probability densities and cumulative probabilities at arbitrary x values can be obtained, along with arbitrary quantiles. The hermite_estimator object only maintains a small and fixed number of coefficients and thus uses minimal memory. The syntax to calculate probability densities, cumulative probabilities and quantiles in the univariate setting is presented below.

7.1 Standard syntax

observations <- rlogis(n=2000)
hermite_est <- hermite_estimator(N=10, standardize=TRUE, 
                                 observations = observations)

x <- seq(-15,15,0.1)
pdf_est <- dens(hermite_est,x)
cdf_est <- cum_prob(hermite_est,x)

p <- seq(0.05,1,0.05)
quantile_est <- quant(hermite_est,p)

7.2 Piped syntax

observations <- rlogis(n=2000)
hermite_est <- hermite_estimator(N=10, standardize=TRUE, 
                                 observations = observations)

x <- seq(-15,15,0.1)
pdf_est <- hermite_est %>% dens(x)
cdf_est <- hermite_est %>% cum_prob(x)

p <- seq(0.05,0.95,0.05)
quantile_est <- hermite_est %>% quant(p)
actual_pdf <- dlogis(x)
actual_cdf <- plogis(x)
df_pdf_cdf <- data.frame(x,pdf_est,cdf_est,actual_pdf,actual_cdf)

actual_quantiles <- qlogis(p)
df_quant <- data.frame(p,quantile_est,actual_quantiles)

7.3 Comparing Estimated versus Actual

ggplot(df_pdf_cdf,aes(x=x)) + geom_line(aes(y=pdf_est, colour="Estimated")) +
  geom_line(aes(y=actual_pdf, colour="Actual")) +
  scale_colour_manual("", 
                      breaks = c("Estimated", "Actual"),
                      values = c("blue", "black")) + ylab("Probability Density")

ggplot(df_pdf_cdf,aes(x=x)) + geom_line(aes(y=cdf_est, colour="Estimated")) +
  geom_line(aes(y=actual_cdf, colour="Actual")) +
  scale_colour_manual("", 
                      breaks = c("Estimated", "Actual"),
                      values = c("blue", "black")) +
  ylab("Cumulative Probability")

ggplot(df_quant,aes(x=actual_quantiles)) + geom_point(aes(y=quantile_est),
                                                      color="blue") +
  geom_abline(slope=1,intercept = 0) +xlab("Theoretical Quantiles") +
  ylab("Estimated Quantiles")

7.4 Convenience functions

Note that there are also generic methods facilitating summarizing and plotting univariate densities and distribution functions as illustrated below.

h_dens <- density(hermite_est)
print(h_dens)
#>  x[1:99] = -4.3577, -3.8235, -3.471,  ...,  4.101, 5.0983
#>  d[1:99] = 0.013719, 0.024207, 0.032781,  ..., 0.012633, 0.0081853
plot(h_dens)


h_cdf <- hcdf(hermite_est)
print(h_cdf)
#>  x[1:99] = -4.3577, -3.8235, -3.471,  ...,  4.101, 5.0983
#>  p[1:99] = 0.0099998, 0.019999, 0.029999,  ..., 0.97922, 0.98922
plot(h_cdf)

summary(h_cdf)
#> Hermite CDF estimates:
#> 2000 observations
#> 99 evaluation points
#>  x[1:99] = -4.3577, -3.8235, -3.471,  ...,  4.101, 5.0983
#>  p[1:99] = 0.0099998, 0.019999, 0.029999,  ..., 0.97922, 0.98922

Finally there are the following convenience functions providing familiar syntax to the ordinary R functions.

quantile(hermite_est)
#> [1] -18.16543137  -1.07927548   0.02118128   1.09551446   9.08612013

median(hermite_est)
#> [1] 0.02118128

IQR(hermite_est)
#> [1] 2.17479

8 Estimate bivariate pdf, cdf and nonparametric correlation

The aforementioned suitability of Hermite series based estimators in sequential and one-pass batch estimation settings extends to the bivariate case. Probability densities and cumulative probabilities can be obtained at arbitrary points. The syntax to calculate probability densities and
cumulative probabilities along with the Spearman and Kendall correlation coefficients in the bivariate setting is presented below.

8.1 Standard syntax

# Prepare bivariate normal data
sig_x <- 1
sig_y <- 1
num_obs <- 4000
rho <- 0.5
observations_mat <- mvtnorm::rmvnorm(n=num_obs,mean=rep(0,2),
          sigma = matrix(c(sig_x^2,rho*sig_x*sig_y,rho*sig_x*sig_y,sig_y^2), 
          nrow=2,ncol=2, byrow = TRUE))

hermite_est <- hermite_estimator(N = 30, standardize = TRUE, 
                                 est_type = "bivariate", 
                                 observations = observations_mat) 
vals <- seq(-5,5,by=0.25)
x_grid <- as.matrix(expand.grid(X=vals, Y=vals))
pdf_est <- dens(hermite_est,x_grid)
cdf_est <- cum_prob(hermite_est,x_grid)
spear_est <- spearmans(hermite_est)
kendall_est <- kendall(hermite_est)

8.2 Piped syntax

sig_x <- 1
sig_y <- 1
num_obs <- 4000
rho <- 0.5
observations_mat <- mvtnorm::rmvnorm(n=num_obs,mean=rep(0,2),
        sigma = matrix(c(sig_x^2,rho*sig_x*sig_y,rho*sig_x*sig_y,sig_y^2), 
          nrow=2, ncol=2, byrow = TRUE))

hermite_est <- hermite_estimator(N = 30, standardize = TRUE, 
                                 est_type = "bivariate", 
                                 observations = observations_mat) 

vals <- seq(-5,5,by=0.25)
x_grid <- as.matrix(expand.grid(X=vals, Y=vals))
pdf_est <- hermite_est %>% dens(x_grid, clipped = TRUE)
cdf_est <- hermite_est %>% cum_prob(x_grid, clipped = TRUE)
spear_est <- hermite_est %>% spearmans()
kendall_est <- hermite_est %>% kendall()
actual_pdf <-mvtnorm::dmvnorm(x_grid,mean=rep(0,2),
            sigma = matrix(c(sig_x^2,rho*sig_x*sig_y,rho*sig_x*sig_y,sig_y^2), 
                           nrow=2,ncol=2, byrow = TRUE))
actual_cdf <- rep(NA,nrow(x_grid))
for (row_idx in seq_len(nrow(x_grid))) {
  actual_cdf[row_idx] <-  mvtnorm::pmvnorm(lower = c(-Inf,-Inf),
    upper=as.numeric(x_grid[row_idx,]),mean=rep(0,2),sigma = matrix(c(sig_x^2, 
        rho*sig_x*sig_y,rho*sig_x*sig_y,sig_y^2), nrow=2,ncol=2,byrow = TRUE))
}
actual_spearmans <- cor(observations_mat,method = "spearman")[1,2]
actual_kendall <- cor(observations_mat,method = "kendall")[1,2]
df_pdf_cdf <- data.frame(x_grid,pdf_est,cdf_est,actual_pdf,actual_cdf)

8.3 Comparing Estimated versus Actual

p1 <- ggplot(df_pdf_cdf) + geom_tile(aes(X, Y, fill= actual_pdf)) +
  scale_fill_continuous_sequential(palette="Oslo",
                                   breaks=seq(0,.2,by=.05),
                                   limits=c(0,.2))

p2 <- ggplot(df_pdf_cdf) + geom_tile(aes(X, Y, fill= pdf_est)) +
  scale_fill_continuous_sequential(palette="Oslo",
                                   breaks=seq(0,.2,by=.05),
                                   limits=c(0,.2))

p1+ ggtitle("Actual PDF")+ theme(legend.title = element_blank()) + p2 +
  ggtitle("Estimated PDF") +theme(legend.title = element_blank()) +
  plot_layout(guides = 'collect')

p1 <- ggplot(df_pdf_cdf) + geom_tile(aes(X, Y, fill= actual_cdf)) +
  scale_fill_continuous_sequential(palette="Oslo",
                       breaks=seq(0,1,by=.2),
                       limits=c(0,1))

p2 <- ggplot(df_pdf_cdf) + geom_tile(aes(X, Y, fill= cdf_est)) +
  scale_fill_continuous_sequential(palette="Oslo",
                                   breaks=seq(0,1,by=.2),
                                   limits=c(0,1))

p1+ ggtitle("Actual CDF") + theme(legend.title = element_blank()) + p2 +
  ggtitle("Estimated CDF") + theme(legend.title = element_blank())+
  plot_layout(guides = 'collect')

Spearman’s correlation coefficient results:

Spearman’s Correlation
Actual 0.453
Estimated 0.447

Kendall correlation coefficient results:

Kendall Correlation
Actual 0.312
Estimated 0.308

8.4 Convenience functions

Note that there are also generic methods facilitating summarizing and plotting bivariate densities and distribution functions as illustrated below.

h_dens <- density(hermite_est)
print(h_dens)
#>        x1      x2        d
#> 1 -1.6335 -1.6229 0.031228
#> 2 -1.6335 -1.2602 0.031581
#> 3 -1.6335 -1.0168 0.032307
#> ...
#>                           
#> 360 1.6459 1.2849 0.041813
#> 361 1.6459 1.6703 0.031762
plot(h_dens)


h_cdf <- hcdf(hermite_est)
print(h_cdf)
#>        x1      x2        p
#> 1 -1.6335 -1.6229 0.012745
#> 2 -1.6335 -1.2602 0.020137
#> 3 -1.6335 -1.0168 0.025357
#> ...
#>                          
#> 360 1.6459 1.2849 0.87033
#> 361 1.6459 1.6703 0.91400
plot(h_cdf)

summary(h_cdf)
#> Hermite CDF estimates:
#> 4000 observations
#> 361 evaluation points
#>        x1      x2        p
#> 1 -1.6335 -1.6229 0.012745
#> 2 -1.6335 -1.2602 0.020137
#> 3 -1.6335 -1.0168 0.025357
#> ...
#>                          
#> 360 1.6459 1.2849 0.87033
#> 361 1.6459 1.6703 0.91400

We also provide a wrapper for the cor function of R with two new methods, namely “hermite.spearman” and “hermite.kendall” as below.

cor(observations_mat,method="hermite.spearman")
#>           [,1]      [,2]
#> [1,] 1.0000000 0.4473602
#> [2,] 0.4473602 1.0000000

cor(observations_mat,method="hermite.kendall")
#>           [,1]      [,2]
#> [1,] 1.0000000 0.3078062
#> [2,] 0.3078062 1.0000000

9 Applying to univariate grouped data (batch setting)

A useful application of the hermite_estimator class is to obtain pdf, cdf and quantile function estimates on groups of data as illustrated in the example below. The fact that the memory usage of hermite_estimator objects is small and fixed irrespective of group size and that batch estimation is fast, implies that it is practical to perform estimation on a very large number of groups.

9.1 Using data.table

We first present the example using the data.table package.

# Prepare Test Data
test_data <- data.frame()
for (i in seq_len(5)) {
  exponential_data <- rexp(n=1000)
  logistic_data <- rlogis(n=1000)
  logn_data <- rlnorm(n=1000)
  test_data <- rbind(test_data,data.frame(dist_name=rep("exp",
                length(exponential_data)),idx=i,observations=exponential_data))
  test_data <- rbind(test_data,data.frame(dist_name=rep("logis",
                      length(logistic_data)),idx=i,observations=logistic_data))
  test_data <- rbind(test_data,data.frame(dist_name=rep("lnorm",
                              length(logn_data)),idx=i,observations=logn_data))
}
setDT(test_data)
# Group observations by distribution and idx and create Hermite estimators
estimates <- test_data[,.(herm_est = list(hermite_estimator(N=10,
             standardize = TRUE, observations = observations))),
             by=.(dist_name,idx)]
estimates
#>     dist_name   idx                       herm_est
#>        <char> <int>                         <list>
#>  1:       exp     1 <hermite_estimator_univar[10]>
#>  2:     logis     1 <hermite_estimator_univar[10]>
#>  3:     lnorm     1 <hermite_estimator_univar[10]>
#>  4:       exp     2 <hermite_estimator_univar[10]>
#>  5:     logis     2 <hermite_estimator_univar[10]>
#>  6:     lnorm     2 <hermite_estimator_univar[10]>
#>  7:       exp     3 <hermite_estimator_univar[10]>
#>  8:     logis     3 <hermite_estimator_univar[10]>
#>  9:     lnorm     3 <hermite_estimator_univar[10]>
#> 10:       exp     4 <hermite_estimator_univar[10]>
#> 11:     logis     4 <hermite_estimator_univar[10]>
#> 12:     lnorm     4 <hermite_estimator_univar[10]>
#> 13:       exp     5 <hermite_estimator_univar[10]>
#> 14:     logis     5 <hermite_estimator_univar[10]>
#> 15:     lnorm     5 <hermite_estimator_univar[10]>
# Group observations by distribution and merge Hermite estimators
merged_estimates <- estimates[,.(herm_comb = list(merge_hermite(herm_est))),
                                by=.(dist_name)]
merged_estimates
#>    dist_name                      herm_comb
#>       <char>                         <list>
#> 1:       exp <hermite_estimator_univar[10]>
#> 2:     logis <hermite_estimator_univar[10]>
#> 3:     lnorm <hermite_estimator_univar[10]>
# Estimate probability densities, cumulative probabilities and quantiles
dens_vals <- merged_estimates[,.(dens_est = list(dens(herm_comb[[1]],
                                            c(0.5,1,1.5,2)))),by=.(dist_name)]
cum_prob_vals <- merged_estimates[,.(cum_prob_est = list(cum_prob(herm_comb[[1]]
                                            ,c(0.5,1,1.5,2)))),by=.(dist_name)]
quantile_vals <- merged_estimates[,.(quantile_est = list(quant(herm_comb[[1]],
                                            c(0.25,0.5,0.75)))),by=.(dist_name)]

9.2 Using dplyr

The same illustrative example is presented using the dplyr package instead of the data.table package below.

# Prepare Test Data
test_data <- data.frame()
for (i in seq_len(5)) {
  exponential_data <- rexp(n=1000)
  logistic_data <- rlogis(n=1000)
  logn_data <- rlnorm(n=1000)
  test_data <- rbind(test_data,data.frame(dist_name=rep("exp",
                length(exponential_data)),idx=i,observations=exponential_data))
  test_data <- rbind(test_data,data.frame(dist_name=rep("logis",
                      length(logistic_data)),idx=i,observations=logistic_data))
  test_data <- rbind(test_data,data.frame(dist_name=rep("lnorm",
                              length(logn_data)),idx=i,observations=logn_data))
}
# Group observations by distribution and idx and create Hermite estimators
estimates <- test_data %>% group_by(dist_name,idx) %>% summarise(herm_est = 
list(hermite_estimator(N=10,standardize = TRUE, observations = observations)))
#> `summarise()` has grouped output by 'dist_name'. You can override using the
#> `.groups` argument.
estimates
#> # A tibble: 15 × 3
#> # Groups:   dist_name [3]
#>    dist_name   idx herm_est       
#>    <chr>     <int> <list>         
#>  1 exp           1 <hrmt_st_ [10]>
#>  2 exp           2 <hrmt_st_ [10]>
#>  3 exp           3 <hrmt_st_ [10]>
#>  4 exp           4 <hrmt_st_ [10]>
#>  5 exp           5 <hrmt_st_ [10]>
#>  6 lnorm         1 <hrmt_st_ [10]>
#>  7 lnorm         2 <hrmt_st_ [10]>
#>  8 lnorm         3 <hrmt_st_ [10]>
#>  9 lnorm         4 <hrmt_st_ [10]>
#> 10 lnorm         5 <hrmt_st_ [10]>
#> 11 logis         1 <hrmt_st_ [10]>
#> 12 logis         2 <hrmt_st_ [10]>
#> 13 logis         3 <hrmt_st_ [10]>
#> 14 logis         4 <hrmt_st_ [10]>
#> 15 logis         5 <hrmt_st_ [10]>
# Group observations by distribution and merge Hermite estimators
merged_estimates <- estimates %>% group_by(dist_name) %>% summarise(herm_comb
                                              = list(merge_hermite(herm_est)))
merged_estimates
#> # A tibble: 3 × 2
#>   dist_name herm_comb      
#>   <chr>     <list>         
#> 1 exp       <hrmt_st_ [10]>
#> 2 lnorm     <hrmt_st_ [10]>
#> 3 logis     <hrmt_st_ [10]>
# Estimate probability densities, cumulative probabilities and quantiles
dens_vals <- merged_estimates %>%
  rowwise() %>% mutate(dens_est = list(dens(herm_comb,c(0.5,1,1.5,2))))
cum_prob_vals <- merged_estimates %>%
  rowwise() %>% mutate(cum_prob_est = list(cum_prob(herm_comb,c(0.5,1,1.5,2))))
quantile_vals <- merged_estimates %>%
  rowwise() %>% mutate(quantile_est = list(quant(herm_comb,c(0.25,0.5,0.75))))

9.3 Checking the results (calculate mean absolute error)

# Compute Mean Absolute Error
dens_vals <- dens_vals %>%
  rowwise() %>% mutate(dens_actual = list(do.call(paste0("d",dist_name),
                list(c(0.5,1,1.5,2))))) %>% mutate(mean_abs_error_density =
                                              mean(abs(dens_est-dens_actual)))
cum_prob_vals <- cum_prob_vals %>%
  rowwise() %>% mutate(cum_prob_actual = list(do.call(paste0("p",dist_name),
                list(c(0.5,1,1.5,2)))))%>% mutate(mean_abs_error_cum_prob = 
                                      mean(abs(cum_prob_est-cum_prob_actual)))
quantile_vals <- quantile_vals %>%
  rowwise() %>% mutate(quantile_actual= list(do.call(paste0("q",dist_name),
            list(c(0.25,0.5,0.75)))))%>% mutate(mean_abs_error_quantiles = 
                                    mean(abs(quantile_est-quantile_actual)))
mean_abs_error_summary <- data.frame(dist_name=dens_vals$dist_name, 
  mean_abs_error_density=dens_vals$mean_abs_error_density, 
  mean_abs_error_cum_prob=cum_prob_vals$mean_abs_error_cum_prob,
  mean_abs_error_quantiles=quantile_vals$mean_abs_error_quantiles)
datatable(mean_abs_error_summary) %>% formatRound(columns = 
  c("mean_abs_error_density","mean_abs_error_cum_prob", 
    "mean_abs_error_quantiles"),digits = 3)

10 Applying to stationary data (sequential setting)

10.1 Univariate Example

Another useful application of the hermite_estimator class is to obtain pdf, cdf and quantile function estimates on streaming data. The speed of estimation allows the pdf, cdf and quantile functions to be estimated in real time. We illustrate this below for cdf and quantile estimation with a sample Shiny application. We reiterate that the particular usefulness is that the full pdf, cdf and quantile functions are updated in real time. Thus, any arbitrary quantile can be evaluated at any point in time. We include a stub for reading streaming data that generates micro-batches of standard exponential i.i.d. random data. This stub can easily be swapped out for a method reading micro-batches from a Kafka topic or similar.

The Shiny sample code below can be pasted into a single app.R file and run directly.

# Not Run. Copy and paste into app.R and run.
library(shiny)
library(hermiter)
library(ggplot2)
library(magrittr)

ui <- fluidPage(
    titlePanel("Streaming Statistics Analysis Example: Exponential 
               i.i.d. stream"),
    sidebarLayout(
        sidebarPanel(
            sliderInput("percentile", "Percentile:",
                        min = 0.01, max = 0.99,
                        value = 0.5, step = 0.01)
        ),
        mainPanel(
           plotOutput("plot"),
           textOutput("quantile_text")
        )
    )
)

server <- function(input, output) {
    values <- reactiveValues(hermite_est = 
                                 hermite_estimator(N = 10, standardize = TRUE))
    x <- seq(-15, 15, 0.1)
    # Note that the stub below could be replaced with code that reads streaming 
    # data from various sources, Kafka etc.  
    read_stream_stub_micro_batch <- reactive({
        invalidateLater(1000)
        new_observation <- rexp(10)
        return(new_observation)
    })
    updated_cdf_calc <- reactive({
        micro_batch <- read_stream_stub_micro_batch()
        for (idx in seq_along(micro_batch)) {
            values[["hermite_est"]] <- isolate(values[["hermite_est"]]) %>%
                update_sequential(micro_batch[idx])
        }
        cdf_est <- isolate(values[["hermite_est"]]) %>%
            cum_prob(x, clipped = TRUE)
        df_cdf <- data.frame(x, cdf_est)
        return(df_cdf)
    })
    updated_quantile_calc <- reactive({
        values[["hermite_est"]]  %>% quant(input$percentile)
    })
    output$plot <- renderPlot({
        ggplot(updated_cdf_calc(), aes(x = x)) + geom_line(aes(y = cdf_est)) +
            ylab("Cumulative Probability")
    }
    )
    output$quantile_text <- renderText({ 
        return(paste(input$percentile * 100, "th Percentile:", 
                     round(updated_quantile_calc(), 2)))
    })
}
shinyApp(ui = ui, server = server)