I spent some time this weekend playing with a couple interesting new R packages that should help with some of the difficulty manipulating data with the base packages. Getting data into a format appropriate for plotting or running statistical models often seems to take more time than anything else, and the process can be very frustrating because of seemingly non-sensical error messages from R.

R guru Hadley Wickham has written some of the best packages for data
manipulation (`reshape2`, `plyr`) and plotting (`ggplot2`). He’s got a
new pair of packages (`tidyr` and `dplyr`) and a theory of getting data into
the proper format (http://vita.had.co.nz/papers/tidy-data.pdf) that look very
promising.

A couple other tools I wanted to look at include a new way of piping data from
one operation to another (`magrittr`), which is a basic part of the Unix
philosophy of having many tools that do one thing well and stringing them
together to do neat things, and an interactive graphic package, `ggvis`, that
should be really good for data investigation.

For this investigation, I’m looking at the names of our dogs, past and present,
and how popular they are as people names in the United States. This data comes
from the Social Security Administration and is available in the R package
`babynames`.

```
install.packages("babynames")
library(babynames)
head(babynames)
Source: local data frame [6 x 5]
year sex name n prop
1 1880 F Mary 7065 0.07238359
2 1880 F Anna 2604 0.02667896
3 1880 F Emma 2003 0.02052149
4 1880 F Elizabeth 1939 0.01986579
5 1880 F Minnie 1746 0.01788843
6 1880 F Margaret 1578 0.01616720
```

The data has the number of registrations (`n`) and proportion of total
registrations (`prop`) for each year from 1880 through 2013 for all name and
sex combinations.

What I want to see is how popular out dog names are. All of our dogs have been adopted, but some we chose names for them (Nika, Piper, Kiva), and the rest came with names we didn’t change (Deuce, Buddy, Koidern, Lennier, Martin, Monte and the second Piper). Dog mushers often choose a theme for a litter of puppies, which accounts for some of the unusual names (Deuce came from a litter of classic cars, Lennier came from a litter of Babylon 5 character names, Koidern from a litter of Yukon River tributary names).

So what I want to do is subset the babynames database to just the names of our dogs, combine male and female names together, and plot the popularity of these names over time.

Here’s how that’s done using the `magrittr` pipe operator (`%>%`):

```
library(dplyr)
library(magrittr)
dog_names <- babynames %>%
filter(name %in% c("Nika", "Piper", "Buddy", "Koidern", "Deuce",
"Kiva", "Lennier", "Martin", "Monte")) %>%
group_by(year, name) %>%
summarise(prop=sum(prop)) %>%
transform(name=factor(name)) %>%
ungroup() %>%
arrange(name, year)
```

We are assigning the result of all the pipes to `dog_names`. In order, we
take the `babynames` data set and filter it by our dog’s names. Then we group
by year and name, and summarize the proportion values by sum. At this point we
have data with just our dog’s names, with the proportions of both male and
female baby names combined. Next, we convert the name variable to a factor,
remove the grouping, and sort by name and year.

Here’s what it looks like now:

```
> head(dog_names)
year name prop
1 1893 Buddy 4.130866e-05
2 1894 Buddy 5.604573e-05
3 1896 Buddy 8.522045e-05
4 1898 Buddy 3.784782e-05
5 1899 Buddy 4.340353e-05
6 1900 Buddy 6.167015e-05
```

Now we’ve got tidy filtered and sorted data, so let’s plot it. I’ve been using
`ggplot2` for many years, and I think it’s the best way to produce publication
quality figures. But usually you want to do some investigation of the data
before doing that, and doing this in `ggplot2` involves many cycles of code
manipulation, plotting, viewing in order to see what you’ve got and how you want
the final version to look.

`ggvis` is a new package that displays data interactively in a web browser.
It also supports the pipe operator, so you can pipe the data directly into the
plotting routine. It’s somewhat similar to `ggplot2`, but has some new
conventions that are required in order to handle interactivity. Here’s a plot
of my dog names data. The first part is the same as before, but I’m piping the
result directly into `ggivs`.

```
library(ggvis)
babynames %>%
filter(name %in% c("Nika", "Piper", "Buddy", "Koidern", "Deuce",
"Kiva", "Lennier", "Martin", "Monte")) %>%
group_by(year, name) %>%
summarise(prop=sum(prop)) %>%
transform(name=factor(name)) %>%
ungroup() %>%
arrange(name, year) %>%
ggvis(~year, ~prop, stroke=~name, fill=~name) %>%
# layer_lines(strokeWidth:=2) %>%
layer_points(size:=15) %>%
add_axis("x", title="Year", format="####") %>%
add_axis("y", title="Proportion of total names", title_offset=50) %>%
add_legend(c("stroke", "fill"), title="Name")
```

Typically, I prefer to include lines and points in a timeseries plot like this,
but I couldn't get `ggvis` to color the lines and the points without some very
strange fill artifacts.

Here’s what I’d consider to be a high quality version of this, generated with
`ggplot`:

```
library(ggplot2)
library(scales)
q <- ggplot(data=dog_names, aes(x=year, y=prop, colour=name)) +
geom_point(size=1.75) +
geom_line() +
theme_bw() +
scale_colour_brewer(palette="Set1") +
scale_x_continuous(name="Year", breaks=pretty_breaks(n=10)) +
scale_y_continuous(name="Proportion of total names", breaks=pretty_breaks(n=10))
rescale <- 0.50
svg("dog_names_ggplot2.svg", height=9*rescale, width=16*rescale)
print(q)
dev.off()
```

I think the two plots are pretty similar, and I’m impressed with how good the
`ggvis` plot looks and how similar the language is to `ggplot2`. And I
really like the pipe operator compared with a long list of individual
statements or the way you add things together with `ggplot2`.

Both plots suffer from having too many groups (seven), which means it becomes
difficult to interpret the colors on the plot. Choosing a good palette is key
to this, and is one of those parts of figure production that can really take a
long time. I don’t think my choices in the `ggplot2` version is optimal, but
I got tired of looking. The other problem is the collection of dog names with
very low proportions among human babies. Because they’re all overlapping near
the axis, this data is obscured. Both problems could be solved by stacking two
plots on top of each other, one with the more popular names (Martin, Piper,
Buddy and Monte) and one with the less popular ones (Deuce, Kiva, Nika) using
different scales for the proportion axis.

What does the plot show? Among our dog’s names, Martin was the most popular, but it’s popularity has been declining since the 60s, and the name Piper has been increasing since 2000. Both Monte and Buddy were popular in the past, but have declined to low levels recently.

For reference, here are the number of babies in 2013 that were given names matching those of our dogs:

```
babynames %>%
filter(name %in% c("Nika", "Piper", "Buddy", "Koidern", "Deuce",
"Kiva", "Lennier", "Martin", "Monte")
& year==2013) %>%
group_by(year, name) %>%
summarise(n=sum(n)) %>%
transform(name=factor(name)) %>%
ungroup() %>%
arrange(desc(n))
year name n
1 2013 Piper 3166
2 2013 Martin 1330
3 2013 Monte 81
4 2013 Nika 67
5 2013 Buddy 21
6 2013 Kiva 18
7 2013 Deuce 5
```

Following up on my previous post, I tried the regression approach for predicting future snow depth from current values. As you recall, I produced a plot that showed how much snow we’ve had on the ground on each date at the Fairbanks Airport between 1917 and 2013. These boxplots gave us an idea of what a normal snow depth looks like on each date, but couldn’t really tell us much about what we might expect for snow depth for the rest of the winter.

# Regression

I ran a linear regression analysis looking at how snow depth on November 8th relates to snow depth on November 27th and December 25th of the same year. Here’s the SQL:

```
SELECT * FROM (
SELECT extract(year from dte) AS year,
max(CASE WHEN to_char(dte, 'mm-dd') = '11-08'
THEN round(snwd_mm/25.4, 1)
ELSE NULL END) AS nov_8,
max(CASE WHEN to_char(dte, 'mm-dd') = '11-27'
THEN round(snwd_mm/25.4, 1)
ELSE NULL END) AS nov_27,
max(CASE WHEN to_char(dte, 'mm-dd') = '12-15'
THEN round(snwd_mm/25.4, 1)
ELSE NULL END) AS dec_25
FROM ghcnd_pivot
WHERE station_name = 'FAIRBANKS INTL AP'
AND snwd_mm IS NOT NULL
GROUP BY extract(year from dte)
ORDER BY year
) AS sub
WHERE nov_8 IS NOT NULL
AND nov_27 IS NOT NULL
AND dec_25 IS NOT NULL;
```

I’m grouping on year, then grabbing the snow depth for the three dates of interest. I would have liked to include dates in January and February in order to see how the relationship weakens as the winter progresses, but that’s a lot more complicated because then we are comparing the dates from one year to the next and the grouping I used in the query above wouldn’t work.

One note on this analysis: linear regression has a bunch of assumptions that need to be met before considering the analysis to be valid. One of these assumptions is that observations are independent from one another, which is problematic in this case because snow depth is a cumulative statistic; the depth tomorrow is necessarily related to the depth of the snow today (snow depth tomorrow = snow depth today + snowfall). Whether it’s necessarily related to the depth of the snow a month from now is less certain, and I’m making the possibly dubious assumption that autocorrelation disappears when the time interval between observations is longer than a few weeks.

# Results

Here are the results comparing the snow depth on November 8th to November 27th:

```
> reg <- lm(data=results, nov_27 ~ nov_8)
> summary(reg)
Call:
lm(formula = nov_27 ~ nov_8, data = results)
Residuals:
Min 1Q Median 3Q Max
-8.7132 -3.0490 -0.6063 1.7258 23.8403
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 3.1635 0.9707 3.259 0.0016 **
nov_8 1.1107 0.1420 7.820 1.15e-11 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 4.775 on 87 degrees of freedom
Multiple R-squared: 0.4128, Adjusted R-squared: 0.406
F-statistic: 61.16 on 1 and 87 DF, p-value: 1.146e-11
```

And between November 8th and December 25th:

```
> reg <- lm(data=results, dec_25 ~ nov_8)
> summary(reg)
Call:
lm(formula = dec_25 ~ nov_8, data = results)
Residuals:
Min 1Q Median 3Q Max
-10.209 -3.195 -1.195 2.781 10.791
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 6.2227 0.8723 7.133 2.75e-10 ***
nov_8 0.9965 0.1276 7.807 1.22e-11 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 4.292 on 87 degrees of freedom
Multiple R-squared: 0.412, Adjusted R-squared: 0.4052
F-statistic: 60.95 on 1 and 87 DF, p-value: 1.219e-11
```

Both regressions are very similar. The coefficients and the overall model are
both very significant, and the *R²* value indicates that in each case, the
snow depth on November 8th explains about 40% of the variation in the snow depth
on the later date. The amount of variation explained hardly changes at all,
despite almost a month difference between the two analyses.

Here's a plot of the relationship between today’s date and Christmas (PDF version)

The blue line is the linear regression model.

# Conclusions

For 2014, we’ve got 2 inches of snow on the ground on November 8th. The models predict we’ll have 5.4 inches on November 27th and 8 inches on December 25th. That isn’t great, but keep in mind that even though the relationship is quite strong, it explains less than half of the variation in the data, which means that it’s quite possible we will have a lot more, or less. Looking back at the plot, you can see that for all the years where we had two inches of snow on November 8th, we had between five and fifteen inches of snow in that same year on December 25th. I’m certainly hoping we’re closer to fifteen.

Winter started off very early this year with the first snow falling on October 4th and 5th, setting a two inch base several weeks earlier than normal. Since then, we’ve had only two days with more than a trace of snow.

This seems to be a common pattern in Fairbanks. After the first snowfall and the establishment of a thin snowpack on the ground, we all get excited for winter and expect the early snow to continue to build, filling the holes in the trails and starting the skiing, mushing and winter fat biking season. Then, nothing.

# Analysis

I decided to take a quick look at the pattern of snow depth at the Fairbanks Airport station to see how uncommon it is to only have two inches of snow this late in the winter (November 8th at this writing). The plot below shows all of the snow depth data between November and April for the airport station, displayed as box and whisker plots.

Here’s the SQL:

```
SELECT extract(year from dte) AS year,
extract(month from dte) AS month,
to_char(dte, 'MM-DD') AS mmdd,
round(snwd_mm/25.4, 1) AS inches
FROM ghcnd_pivot
WHERE station_name = 'FAIRBANKS INTL AP'
AND snwd_mm IS NOT NULL
AND (extract(month from dte) BETWEEN 11 AND 12
OR extract(month from dte) BETWEEN 1 AND 4);
```

If you’re interested in the code that produces the plot, it’s at the bottom of the post. If the plot doesn’t show up in your browser or you want a copy for printing, here’s a link to the PDF version.

# Box and whisker plots

For those unfamiliar with these, they’re a good way to evaluate the range of data grouped by some categorical variable (date, in our case) along with details about the expected values and possible extremes. The upper and lower limit of each box show the ranges where 25—75% of the data fall, meaning that half of all observed values are within this box for each date. For example, on today’s date, November 8th, half of all snow depth values for the period in question fell between four and eight inches. Our current snow depth of two inches falls below this range, so we can say that only having two inches of snow on the ground happens in less than 25% of the time.

The horizontal line near the middle of the box is the median of all observations for that date. Median is shown instead of average / mean because extreme values can skew the mean, so a median will often be more representative of the most likely value. For today’s date, the median snow depth is five inches. That’s what we’d expect to see on the ground now.

The vertical lines extending above and below the boxes show the points that are within 1.5 times the range of the boxes. These lines represent the values from the data outside the most likely, but not very unusual. If you scan across the November to December portion of the plot, you can see that the lower whisker touches zero for most of the period, but starting on December 26th, it rises above zero and doesn’t return until the spring. That means that there have been years where there was no snow on the ground on Christmas. Ouch.

The dots beyond the whiskers are outliers; observations so far from what is normal that they’re exceptional and not likely to be repeated. On this plot, most of these outliers are probably from one or two exceptional years where we got a ton of snow. Some of those outliers are pretty incredible; consider having two and a half feet of snow on the ground at the end of April, for example.

# Conclusion

The conclusion I’d draw from comparing our current snow depth of two inches against the boxplots is that it is somewhat unusual to have this little snow on the ground, but that it’s not exceptional. It wouldn’t be unusual to have no snow on the ground.

Looking forward, we would normally expect to have a foot of snow on the ground
by mid-December, and I’m certainly hoping that happens this year. But despite
the probabilities shown on this plot it can’t say how likely that is when we
*know* that there’s only two inches on the ground now. One downside to boxplots
in an analysis like this is that the independent variable (date) is categorical,
and the plot doesn’t have anything to say about how the values on one day relate
to the values on the next day or any date in the future. One might expect, for
example, that a low snow depth on November 8th means it’s more likely we’ll also
have a low snow depth on December 25th, but this data can’t offer evidence on
that question. It only shows us what each day’s pattern of snow depth is
expected to be on it’s own.

Bayesian analysis, “given a snow depth of two inches on November 8th, what is the likelihood of normal snow depth on December 25th”, might be a good approach for further investigation. Or a more traditional regression analysis examining the relationship between snow depth on one date against snow depth on another.

# Appendix: R Code

```
library(RPostgreSQL)
library(ggplot2)
library(scales)
library(gtable)
# Build plot "table"
make_gt <- function(nd, jf, ma) {
gt1 <- ggplot_gtable(ggplot_build(nd))
gt2 <- ggplot_gtable(ggplot_build(jf))
gt3 <- ggplot_gtable(ggplot_build(ma))
max_width <- unit.pmax(gt1$widths[2:3], gt2$widths[2:3], gt3$widths[2:3])
gt1$widths[2:3] <- max_width
gt2$widths[2:3] <- max_width
gt3$widths[2:3] <- max_width
gt <- gtable(widths = unit(c(11), "in"), heights = unit(c(3, 3, 3), "in"))
gt <- gtable_add_grob(gt, gt1, 1, 1)
gt <- gtable_add_grob(gt, gt2, 2, 1)
gt <- gtable_add_grob(gt, gt3, 3, 1)
gt
}
drv <- dbDriver("PostgreSQL")
con <- dbConnect(drv, dbname="DBNAME")
results <- dbGetQuery(con,
"SELECT extract(year from dte) AS year,
extract(month from dte) AS month,
to_char(dte, 'MM-DD') AS mmdd,
round(snwd_mm/25.4, 1) AS inches
FROM ghcnd_pivot
WHERE station_name = 'FAIRBANKS INTL AP'
AND snwd_mm IS NOT NULL
AND (extract(month from dte) BETWEEN 11 AND 12
OR extract(month from dte) BETWEEN 1 AND 4);")
results$mmdd <- as.factor(results$mmdd)
# NOV DEC
nd <- ggplot(data=subset(results, month == 11 | month == 12), aes(x=mmdd, y=inches)) +
geom_boxplot() +
theme_bw() +
theme(axis.title.x = element_blank()) +
theme(plot.margin = unit(c(1, 1, 0, 0.5), 'lines')) +
# scale_x_discrete(name="Date (mm-dd)") +
scale_y_discrete(name="Snow depth (inches)", breaks=pretty_breaks(n=10)) +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
ggtitle('Snow depth by date, Fairbanks Airport, 1917-2013')
# JAN FEB
jf <- ggplot(data=subset(results, month == 1 | month == 2), aes(x=mmdd, y=inches)) +
geom_boxplot() +
theme_bw() +
theme(axis.title.x = element_blank()) +
theme(plot.margin = unit(c(0, 1, 0, 0.5), 'lines')) +
# scale_x_discrete(name="Date (mm-dd)") +
scale_y_discrete(name="Snow depth (inches)", breaks=pretty_breaks(n=10)) +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) # +
# ggtitle('Snowdepth by date, Fairbanks Airport, 1917-2013')
# MAR APR
ma <- ggplot(data=subset(results, month == 3 | month == 4), aes(x=mmdd, y=inches)) +
geom_boxplot() +
theme_bw() +
theme(plot.margin = unit(c(0, 1, 1, 0.5), 'lines')) +
scale_x_discrete(name="Date (mm-dd)") +
scale_y_discrete(name="Snow depth (inches)", breaks=pretty_breaks(n=10)) +
theme(axis.text.x = element_text(angle = 45, hjust = 1)) # +
# ggtitle('Snowdepth by date, Fairbanks Airport, 1917-2013')
gt <- make_gt(nd, jf, ma)
svg("snowdepth_boxplots.svg", width=11, height=9)
grid.newpage()
grid.draw(gt)
dev.off()
```

It’s the beginning of a new year and time for me to look back at what I learned last year. Rather than a long narrative, let’s focus on the data. The local newspaper did a “community profile” of me this year and it was focused on my curiosity about the world around us and how we can measure and analyze it to better understand our lives. This post is a brief summary of that sort of analysis for my small corner of the world in the year that was 2013.

# Exercise

2013 was the year I decided to, and did, run the Equinox Marathon, so I spent a lot of time running this year and a lot less time bicycling. Since the race, I’ve been having hip problems that have kept me from skiing or running much at all. The roads aren’t cleared well enough to bicycle on them in the winter so I got a fat bike to commute on the trails I’d normally ski.

Here are my totals in tabular form:

type | miles | hours | calories |
---|---|---|---|

Running | 529 | 89 | 61,831 |

Bicycling | 1,018 | 82 | 54,677 |

Skiing | 475 | 81 | 49,815 |

Hiking | 90 | 43 | 18,208 |

TOTAL | 2,113 | 296 | 184,531 |

I spent just about the same amount of time running, bicycling and skiing this year, and much less time hiking around on the trails than in the past. Because of all the running, and my hip injury, I didn’t manage to commute to work with non-motorized transport quite as much this year (55% of work days instead of 63% in 2012), but the exercise totals are all higher.

One new addition this year is a heart rate monitor, which allows me to much more accurately estimate energy consumption than formulas based on the type of activity, speed, and time. Riding my fat bike, it’s pretty clear that this form of travel is so much less efficient than a road bike with smooth tires that it can barely be called “bicycling,” at least in terms of how much energy it takes to travel over a certain distance.

Here’s the equations from Keytel LR, Goedecke JH, Noakes TD, Hiiloskorpi H,
Laukkanen R, van der Merwe L, Lambert EV. 2005. *Prediction of energy
expenditure from heart rate monitoring during submaximal exercise.* J Sports
Sci. 23(3):289-97.

*Male*: ( − 55.0969 + (0.6309

*hr*) + (0.0901

*w*) + (0.2017

*a*))/(4.184)60

*t*

*Female*: ( − 20.4022 + (0.4472

*hr*) − (0.0901

*w*) + (0.074

*a*))/(4.184)60

*t*

where

hr= Heart rate (in beats/minute)w= Weight (in pounds)a= Age (in years)t= Exercise duration time (in hours)

And a SQL function that implements the version for men (to use it, you’d replace
the `nnn`

and `yyyy-mm-dd`

with the appropriate values for you):

```
--- Kcalories burned based on average heart rate and number
--- of hours at that rate.
CREATE OR REPLACE FUNCTION kcal_from_hr(hr numeric, hours numeric)
RETURNS numeric
LANGUAGE plpgsql
AS $$
DECLARE
weight_lb numeric := nnn;
resting_hr numeric := nn;
birthday date := 'yyyy-mm-dd';
resting_kcal numeric;
exercise_kcal numeric;
BEGIN
resting_kcal := ((-55.0969+(0.6309*(resting_hr))+
(0.0901*weight_lb)+
(0.2017*(extract(epoch from now()-birthday)/
(365.242*24*60*60))))/4.184)*60*hours;
exercise_kcal := ((-55.0969+(0.6309*(hr))+
(0.0901*weight_lb)+
(0.2017*(extract(epoch from now()-birthday)/
(365.242*24*60*60))))/4.184)*60*hours;
RETURN exercise_kcal - resting_kcal;
END;
$$;
```

Here’s a graphical comparison of my exercise data over the past four years:

It was a pretty remarkable year, although the drop in exercise this fall is disappointing.

Another way to visualize the 2013 data is in the form of a heatmap, where each block represents a day on the calendar, and the color is how many calories I burned on that day. During the summer you can see my long runs on the weekends showing up in red. Equinox was on September 21st, the last deep red day of the year.

# Weather

2013 was quite remarkable for the number of days where the daily temperature was
dramatically different from the 30-year average. The heatmap below shows each
day in 2013, and the color indicates how many standard deviations that day’s
temperature was from the 30-year average. To put the numbers in perspective,
approximately 95.5% of all observations will fall within two standard deviations
from the mean, and 99.7% will be within three standard deviations. So the very
dark red or dark blue squares on the plot below indicate temperature anomalies
that happen less than 1% of the time. Of course, in a full year, you’d expect
to see a few of these remarkable differences, but 2013 had *a lot* of remarkable
differences.

2013 saw 45 days where the temperature was more than 2 standard deviations from the mean (19 that were colder than normal and 26 that were warmer), something that should only happen 16 days out of a normal year [ 365.25(1 − 0.9545) ]. There were four days ouside of 3 standard deviations from the mean anomaly. Normally there’d only be a single day [ 365.25(1 − 0.9973) ] with such a remarkably cold or warm temperature.

April and most of May were remarkably cold, resulting in many people skiing long past what is normal in Fairbanks. On May first, Fairbanks still had 17 inches of snow on the ground. Late May, almost all of June and the month of October were abnormally warm, including what may be the warmest week on record in Alaska from June 23rd to the 29th. Although it wasn’t exceptional, you can see the brief cold snap preceding and including the Equinox Marathon on September 21st this year. The result was bitter cold temperatures on race day (my hands and feet didn’t get warm until I was climbing Ester Dome Road an hour into the race), as well as an inch or two of snow on most of the trail sections of the course above 1,000 feet.

Most memorable was the ice and wind storm on November 13th and 14th that dumped several inches of snow and instantly freezing rain, followed by record high winds that knocked power out for 14,000 residents of the area, and then a drop in temperatures to colder than ‒20°F. My office didn’t get power restored for four days.

## git

I’m moving more and more of my work into git, which is a distributed revision
control system (or put another way, it’s a system that stores stuff and keeps
track of all the changes). Because it’s distributed, anything I have on my
computer at home can be easily replicated to my computer at work or anywhere
else, and any changes that I make to these files on any system, are easy to
recover anywhere else. And it’s all backed up on the master repository, and
*all* changes are recorded. If I decide I’ve made a mistake, it’s easy to go
back to an earlier version.

Using this sort of system for software code is pretty common, but I’m also using
this for normal text files (the `docs`

repository below), and have
starting moving other things into git such as all my eBooks.

The following figure shows the number of file changes made in three of my
repositories over the course of the year. I don’t know why April was such an
active month for Python, but I clearly did a lot of programming that month. The
large number of file changes during the summer in the `docs`

repository is
because I was keeping my running (and physical therapy) logs in that repository.

# Dog Barn

The dog barn was the big summer project. It’s a seven by eleven foot building with large dog boxes inside that we keep warm. When the temperatures are too cold for the dogs to stay outside, we put them into their boxes in the dog barn and turn the heat up to 40°F. I have a real-time visualization of the conditions inside and outside the barn, and because the whole thing is run with a small Linux computer and Arduino board, I’m able to collect a lot of data about how the barn is performing.

One such analysis will be to see how much heat the dogs produce when they are in the barn. To estimate that, we need a baseline of how much heat we’re adding at various temperatures in order to keep it at temperature. I haven’t collected enough cold temperature data to really see what the relationship looks like, but here’s the pattern so far.

The graph shows the relationship between the temperature differential between the outside and inside of the barn plotted against the percentage of time the heater is on in order to maintain that differential, for all 12-hour periods where the dogs weren’t in the barn and there’s less than four missing observations. I’ve also run a linear and quadratic regression in order to predict how much heat will be required at various temperature differentials.

The two r^{2} values shows how much of the variation in heating is explained
by the temperature differential for the linear and the quadratic regressions. I
know that this isn’t a linear relationship, but that model still fits the data
better than the quadratic model does. It may be that it’s some other form of
non-linear relationship that’s not well expressed by a second order polynomial.

Once we can predict how much heat it *should* take to keep the barn warm at a
particular temperature differential, we can see how much less heat we’re using
when the dogs are in the barn. One complication is that the dogs produce enough
moisture when they are in the barn that we need to ventilate it when they are in
there. So in addition to the additive heating from the dogs themselves, there
will be increased heat losses because we have to keep it better ventilated.

It’ll be an interesting data set.

# Power

Power consumption is a concern now that we’ve set up the dog barn and are keeping it heated with an electric heater. It’s an oil-filled radiator-style heater, and uses around 1,100 Watts when it’s on.

This table shows our overall usage by year for the period we have data.

year | average watts | total KWH |
---|---|---|

2010 | 551 | 4822 |

2011 | 493 | 4318 |

2012 | 433 | 3792 |

2013 | 418 | 3661 |

Our overall energy use continues to go down, which is a little surprising to me, actually, since we eliminated most of the devices known to use a lot electricity (incandescent light bulbs, halogen floodlights) years ago. Despite that, and bringing the dog barn on line in late November, we used less electricity in 2013 than in the prior three years.

Here’s the pattern by month, and year.

The spike in usage in November is a bit concerning, since it’s the highest overall monthly consumption for the past four years. Hopefully this was primarily due to the heavy use of the heater during the final phases of the dog barn construction. December wasn’t a particularly cold month relative to years past, but it’s good to see that our consumption was actually quite low even with the barn heater being on the entire month.

That wraps it up. Have a happy and productive 2014!

My last blog post compared the time for the men who ran both the 2012 Gold Discovery Run and the Equinox Marathon in order to give me an idea of what sort of Equinox finish time I can expect. Here, I’ll do the same thing for the 2012 Santa Claus Half Marathon.

Yesterday I ran the half marathon, finishing in 1:53:08, which is an average pace of 8.63 / 8:38 minutes per mile. I’m recovering from a mild calf strain, so I ran the race very conservatively until I felt like I could trust my legs.

I converted the SportAlaska PDF files the same way as before, and read the data in from the CSV files. Looking at the data, there are a few outliers in this comparison as well. In addition to being ouside of most of the points, they are also times that aren’t close to my expected pace, so are less relevant for predicting my own Equinox finish. Here’s the code to remove them, and perform the linear regression:

```
combined <- combined[!(combined$sc_pace > 11.0 | combined$eq_pace > 14.5),]
model <- lm(eq_pace ~ sc_pace, data=combined)
summary(model)
Call:
lm(formula = eq_pace ~ sc_pace, data = combined)
Residuals:
Min 1Q Median 3Q Max
-1.08263 -0.39018 0.02476 0.30194 1.27824
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) -1.11209 0.61948 -1.795 0.0793 .
sc_pace 1.44310 0.07174 20.115 <2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 0.5692 on 45 degrees of freedom
Multiple R-squared: 0.8999, Adjusted R-squared: 0.8977
F-statistic: 404.6 on 1 and 45 DF, p-value: < 2.2e-16
```

There were fewer male runners in 2012 that ran both Santa Claus and Equinox, but we get similar regression statistics. The model and coefficient are significant, and the variation in Santa Claus pace times explains just under 90% of the variation in Equinox times. That’s pretty good.

Here’s a plot of the results:

As before, the blue line shows the model relationship, and the grey area surrounding it shows the 95% confidence interval around that line. This interval represents the range over which 95% of the expected values should appear. The red line is the 1:1 line. As you’d expect for a race twice as long, all the Equinox pace times are significantly slower than for Santa Claus.

There were fewer similar runners in this data set:

Runner | DOB | Santa Claus | Equinox Time | Equinox Pace |
---|---|---|---|---|

John Scherzer | 1972 | 8:17 | 4:49 | 11:01 |

Greg Newby | 1965 | 8:30 | 5:03 | 11:33 |

Trent Hubbard | 1972 | 8:31 | 4:48 | 11:00 |

This analysis predicts that I should be able to finish Equinox in just under five hours, which is pretty close to what I found when using Gold Discovery times in my last post. The model predicts a pace of 11:20 and an Equinox finish time of four hours and 57 minutes, and these results are within the range of the three similar runners listed above. Since I was running conservatively in the half marathon, and will probably try to do the same for Equinox, five hours seems like a good goal to shoot for.