Tornado_with_rmd.Rmd

---
title: "The one-in-1,000-year tornado and its frequency in 1,000 years"
output: html_document
author: Georg Veh
date: "05 October 2018"
---

```{r , include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

We simulate a time series of 1,000 years. For each year, we
determine independently with a random number generator whether
a 1,000-year tornado was observed or not:


```{r , echo=TRUE}

# Generate 1,000 years of random uniform numbers
y <- runif(1000)

# Count only those numbers >=0.999 (or <0.001)
(obs <- sum(y >= 0.999))

```

Re-run this code several times to confirm differing tornado counts;
set.seed() returns the same random numbers for a given
argument.

We can now create many time series as virtual experiments, and
plot the number of tornado counts from each series to learn more
about their *distribution*:

```{r, echo=TRUE}

# Create 1,000 time series, each with 1,000 years
# and store the result column-wise in a matrix
y_long <- matrix(runif(1e6), ncol = 1000)
# Compute the column sums (1,000-year tornadoes)
obs_long <- colSums(y_long >= 0.999)

```

And here the code how to plot this:

```{r, echo = T}

# Plot tornado counts
plot(table(obs_long), lwd = 10,
main = "Simulated 1,000-year Tornadoes",
col = "hotpink",
xlab = "Tornado count",
ylab = "Frequency",
cex.lab = 1.5, cex.main = 1.5)

```

Now do some changes to see how Github manages this
Maybe a barplot with lightblue bars?

```{r, echo = T}

# Plot tornado counts
barplot(table(obs_long), lwd = 0.8, las = 1,
main = "Simulated 1,000-year Tornadoes",
col = "lightblue",
xlab = "Tornado count",
ylab = "Frequency",
lend = 1,
cex.lab = 0.8, cex.main = 0.8)

```