Predicted Wedding Date

How people's prediction of a friend's wedding compared to the actual wedding date.

Although it was a small data set (N=11), it was interesting to see the distribution between females (who tended to predict the wedding would be sooner) and males.

Party Volume Tracking

Recently I hosted a Bday party for Amy.  I took this opportunity to test out a sensor light system I made.

Although there are limitations, it was still fun to see the data from the party (below).  Enjoy :)

*Data was collected 5 times per second

*Sensor system only had one microphone and was placed on a table by the wall

*Data was smoothed using a running average filter

DC Metro patterns

Using the recent publicly available WMATA (DC's Metro system) ridership data, I generated several network visualizations that highlight many interesting ridership patterns.  For example, you can clearly see that Farragut North, Farragut West, Metro Center, and L'Enfant Plaza are the work destination of many DC commuters.

You can also see that Gallery Place Chinatown, Dupont Circle, U Street, and Clarendon are the central night life locations.

It is also interesting to see the drop off in metro traffic on Sundays compared to Saturdays.

 My class report can be found at: https://wiki.cs.umd.edu/cmsc734_f12/index.php?title=Fong%2C_Allan_-_Metro_Ridership

Wedding day temperature

I wanted to see which weather site was better at forecasting the temperature on our wedding day (6/16/12) in Arlington, VA.  The two sites I monitored were: www.accuweather.com and www.weather.com.

Below is a graph of each day's high/low temperature predictions for 6/16/12.
Note:
The green line is the actual recorded high (82) and low (63) temperatures for 6/16/12.
Weather.com only started updating their predictions on 6/7/12.

The average root mean square of the errors for weather.com and accuweather.com are 2.458 and 4.172 respectively.
Weather.com was better at predicting the high and low temperatures for our wedding day.

Wedding RSVPs

Waiting to get a finalized head count for my wedding was a pretty interesting time.  I wanted to see how quickly it would take to get a finalized head count.

The data below includes only those who were a "Maybe going to the wedding" from an initial survey.

The graphs show how quickly the "Maybe" folks made their decisions.
Families/Couples are considered one unit, one count.
The wedding invites was sent out on 4/18/12 with a requested RSVP by 5/20/12.

The graph below shows the daily count of "Maybe" folk responses.

The graph below is the daily Cumulative Distribution Function of the "Maybe" folk responses.

Responses peak at the beginning of the RSVP phase and in the last week of the RSVP phase.  We received the most responses between 5/20 and 5/22.  Assuming it takes about 1-2 days for the RSVP to be delivered, I'd say that most people responded on-time.  By 5/22, our "Maybe" responses were only 80% complete.  Amy and I followed up with a couple of e-mails helping us finalize our head count shortly after.  Thank you all for your responses and for making this data possible :)

Facebook comments

Last month, Amy and I changed our Facebook status to "Engaged" at the same time: 1415 (2:15pm).
Below is a cumulative distribution plot of Facebook comments on our changed status.
Data was sampled at 30 minute intervals.
















Amy's friends commented much quicker but my page experienced a much more steady inflow of comments.

Bananas

I was curious if there is a way to store bananas to keep them from browning so quickly.
I kept bananas (a) laid flat, (b) in a bag, and (c) suspended for a week.






















Results:
The bananas in the bag kept the longest (apparently oxygen causes bananas to brown).
Bananas laid flat and hung basically browned at the same rate.
Eat your bananas within 6 days of buying them (unless they are kept in a bag).

Bananas laid flat












Bananas in bag












Bananas hung

Fall Gathering Attendance

Attendance pattern at the annual Fall Gathering.

-Data does not include the 3 hosts
-Party was from 7-10pm (at least that's what it said on the invite)
-Peak of the party was between 8:12pm and 8:45pm
-Most everyone arrived before 8:30pm (somewhat linear function)
-People tended to leave in groups (step function)

2011 Summer Movies

This was a good summer at the movies, especially as a superhero fan. I wanted to see how the top box office movie did, not just in the opening weekend but throughout the summer. I only included the movies that reached number 1 status at some point or another between the weeks of 5/8/11 and 9/4/11.

Graph 1 and Graph 2 show that the peaks of summer movie sales were 7/17 (driven by Harry Potter) and 7/3 (driven by Transformers). However, Graph 3 and Graph 4 show that Harry Potter, Transformers, and Green Lantern have the shortest half-life (time it takes for their weekly revenue to decrease by half of its opening earnings). Thor and Super 8 have the longest half-life.

Graph 1 shows the cumulative total of summer movie sales.

Graph 2 shows the totals for each summer movie (note that Captain American and Harry Potter are still going kind of strong).

Graph 3 shows the delay rate of the summer movies.

In Graph 4, the half-life of Harry Potter is 0.7 weeks, Transformers is 0.97 weeks, Green Lantern is 0.76 weeks, Super 8 is 1.5 weeks, and Thor is 1.1 weeks (the x-axis is number weeks from opening day).

Thoughts:

I thought it was interesting how quickly people lost interest in Green Lantern and how quickly it got over shadowed by Cars 2. It was also interesting to see how well Hangover did, especially since it came out only a week after Pirates. Thor and Super 8 held in pretty well (you still see Thor advertisement here and there) considering it didn't make the big bucks. (I'm guessing Thor did well because it was the first superhero movie of the summer and I'm guessing Super 8 did well because it was just something different.) Harry Potter and Transformers was not a surprise but I was kind of happy to see Captain America kick Harry Potter out of the number one spot.

Coffee Hour attendence

The graph below shows the number of people at a GraceDC coffee hour after church (recorded at 5 minute intervals).
It appears that coffee hour is most busy around the 15 minute mark.
Most people leave within 25 minutes.
Those remaining after 30 minutes are typically GraceDC members and they tend to leave coffee hour at a constant rate (linear slope).

Sport Entertainment

It seems like no matter what time of year it is, there is always some kind of major sporting event to watch. So I was curious if this was actually true or just my perception. Below I plotted the typical "Amount of Sport Entertainment" in given year for the 4 major sporting leagues in the US (NBA, NFL, MBL, and NHL).

There is definitely a low in the June/July time frame. Maybe the sports industry does not want to compete with the summer movies.

pre = preseason = .5
r = regular season = 1
post = post season = 1.5
f = finals = 1.5

Jan.1 = first half of January
Jan.2 = second half of January

Birthday Posts

You know how on your b-day, you typically get a ton of Birthday wishes on Facebook?
Well, I was curious if there was a pattern to this madness so I graphed the distribution of facebook birthday messages for each hour on the day of my birthday (by percentage).

There appears to be two peaks around 2PM and 9PM. I wonder if that is suggestive of the times when most people check their facebook.

Salt intake

Inspired by the recent dietary guidelines (http://www.cnpp.usda.gov/dietaryguidelines.htm), I wanted to see how good people are at estimating the amount of salt in food. Below are box-plots of the percentage of one's daily recommended value of salt (based on a 2,000 calorie diet) people THOUGHT were contained in a serving size for the different foods. In addition to what people thought, the green squares represent the ACTUAL percentage of one's daily recommended value of salt.
















It is interesting that, people typically over-estimate the amount of salt included in the foods (except for the bagel and black beans), but at the same time, people thought our daily recommended value of salt is 5.43 grams while in reality it is 2.4 grams or 1 teaspoon.
N = 11

Items include (serving size):
Thomas plain bagel (1 bagel)
Bush's black beans (1/2 cup)
Campbell's tomato soup (1 cup)
Utz wheat n cheddar crackers (1 pkg)
Oscar Mayor bologna (1 slice)
Snyder's sourdough nibblers (16 nibblers)
Peter Pan crunchy peanut butter (2 tbsp)
Planters Lightly salted cashews (1 oz)
Oreo (3 cookies)

Songs during Glee

I like the songs in Glee but not so much the storyline and drama. If I can figure out when the songs typically occur, then I can tune in when there is the greatest probability that there will be a song and not have to sit through the drama. The best chances of catching a song appears to be 11, 27, and 41 minutes into the show.

Average number of song per show: 6.2 songs
Average duration of songs: 1.8 minutes (stdv of 0.6 minutes)
N = 6 shows
I know I need a larger sample size, anyone wanna help me collect data? :)

Overlap in TV channels

I am always curious how TV channels overlapped in terms of similar "themed" shows. I looked at all the shows currently (2010) on for three of my favorite channels (Food Network, Science Channel, and Travel Channel) and mapped the number of what I considered "unique" and "overlapping" shows into a venn-diagram.

The most overlap was between the Food Network and the Travel Channel...probably because traveling and culture are so tightly linked with food. The Food Network also has the most shows, probably an indication of its willingness to explore with different themed shows and ideas.

When babies are born

Graph below shows that most of my friends have birthdays in late fall or in March. There is also an increasing trend in birthdays from May to November. Kind of interesting assuming that most babies are born 9 month after conception :P

Baby Sleep Patterns

Questions:
How do variables, such as number of naps during the day or time of sleep, affect the time Baby wakes up the next day?

Independent variables:
A - Number of naps during the day before
B - Total time (mins) napping the day before
C - Amount of time (mins) awake between last nap and sleep
D - Time of sleep
E - Amount of sleep (mins) that night
F - Amount of time (mins) awake during the night
G - Month

Dependent variable:
Time Baby wakes up

Overall results:
Independent variables are "Okay" at predicting when Baby wakes up. Results are useful to inform but not drive decisions. The effects are not statistically significant and models have mediocre predictive powers, probably due to other unaccounted confounding factors.

ANOVA results/discussions:
-Though, none of the variables had a statistically significant effect, factors C (amount of time awake between last nap and sleep) and F (amount of time awake during the night) tend to have the most effect on when Baby wakes up.
-Results could probably improve if I used time of sleep as a covariate ("future work" hehe).

Classification Tree results/discussions:
-Sleep time before or after 9:22pm has a big effect as to when Baby wakes up.
-If Baby is awake for more than 292.5 mins (4.9 hrs) before sleeping, Baby will tend to wake up later.
-As seem from the second graph below, this classification tree model tends to under-predict when Baby wakes up.



Neural Network results/discussions:
-Factors used in this model are "okay" as predicting when Baby wakes up.
-Test correlation was 0.88 which is pretty good. However, the overall performance was not great (R=0.5) probably due to over-training of the network.

Sudoku Solver

Finished*!!!

Click here:
Sudoku Solver

*This works for almost all Sudoku though there are a few that doesn't work for some reason...blah
**Note the code is really really ugly, I might go back to clean it up or write a neater version (maybe...). and if you have suggestions, please let me know

Food and stories

This is a little experiment I conducted to see how different descriptors of a food effect one's enjoyment and receptiveness of that food. Thank you for all those two participated :)

Both groups were given a little cracker packet, a descriptor, and two short questions that addressed one's enjoyment and receptiveness to the crackers.

Group A (N=10)
Descriptor: Growing up in Taiwan, these rice crackers were special treats for us, especially around b-days.

Group B (N=9)
Descriptor: These rice crackers are light and fluffy with a salty sweet flavor.

Discussion:
-It was difficult to account for many factors such as previous exposures to Asian food and tastes, which probably resulted in the large error bars.
-However, the graph does suggest that although personal enjoyment/taste preference is not effected by the descriptor, one's receptiveness and willingness to try the food again might be increased by a more personal descriptor.
-Need a large N (as always) :P

Actual Driving Times vs Google Driving Times

Data from actual driving times in BLUE.
Google estimated driving times in RED (there was not a with/without traffic distinction for these trips)



Trip A:
Estimated: 10.1 miles in 20 mins


Trip B:
Estimated: 8.6 miles in 17 mins