Tag Archives: statistics

Halloween 2016

Halloween 2016 fell on a Monday school night evening. This was also a Monday social security day – the end of the month when senior citizens come to Kolonia to collect their social security checks and go shopping. That income is important in many families here, and falling on a Monday meant that the Halloween shopping weekend would likely have been negatively impacted. In local parlance, October 29 and 30 were a “broke weekend.”

Tristan and Kisha Halloween 2016
Tristan and Kisha Halloween 2016

The weather was acceptable, only a brief passing light rain shower in Dolihner, otherwise generally dry conditions.

Perhaps the largest factor was that last year Halloween fell on a Saturday night. A weekend with no school the next day.

Whatever the underlying factors, numbers were down year-on-year. Groups are a very roughly estimated with overestimation more likely than under. That said, the front porch saw a drop from 90 groups in 2015 to 79 groups in 2016. Traffic began around 18:35 but by 20:30 no further trick or treaters arrived on the porch.

Halloween group sizes 2015
Halloween group sizes 2015

Note the nine outlying groups in 2015 – groups with more than roughly 15 candy receivers, including one up near 45 and another above 50. The differential in the number of groups is a drop of only eleven. The lack of large groups, however, meant raw numbers of individual candy takers was down more significantly.

Halloween group sizes 2016
Halloween group sizes 2016

The numbers were down even more significantly. The count of candy receivers in 2015 was 590. In 2016 only 416 showed up on the porch, a drop of 174 trick or treaters. Average group size also dropped, primarily a function of the drop in the number of large groups and the absence of any group larger that 35. The household thought that the choice to block cars from driving up the interior road negatively impacted the large group counts. My sense is that the large trucks used to haul the big groups of kids from other parts of the island may not have been as available as they were on a Saturday night last night.

In 2015 the average group size was 6.56 with a standard deviation of 8.90. In 2016 average group size was 5.27 with a standard deviation of 5.50. The median, however, increased from 3 to 4 year-on-year.

We again used the dual bowl system. One twenty-five dollar bag of better candy and a single 330 count bag of Hershey Kisses. Elterina added in three bags of additional small candies that may have added upwards of 90 candies to the Kisses bowl. We would end the evening with candy on hand.

For those who want to play with the raw data, the data is available in a Google Sheets spreadsheet. Analysis was done using Google Sheets with the above charts prepared using the Google Statistics add-in for Google Sheets.

Pacific island dance judging analysis

Last Saturday I assisted with a judging a preliminary round of a Pacific island dance contest. This Saturday I was one of four judges for the final round. Four judges judges seven groups on five criteria. The criteria were modified from last week.

1. Movement. (Late movement, Turn the other way, keep looking at the partner, etc.)
2. Costume
3. Always smile? Ashamed? No singing with the music?
4. Well practice?
5. Performance as a group
Chewing gum and chewing betelnut and spitting: 10 point deduction.

Each criteria was worth ten points. Each group danced two dances, each dance could generate up to 5 of the 10 points in each criteria.

Bring it on girls
Bring it on girls

With four judges, five criteria each worth ten points, in theory there was a maximum of 200 points possible. At the end of the evening the rank order and points for the seven dance groups ranged from 152 to 182 points.

Dancers Sum
Bring it on girls, Pohnlik, Kolonia 182
Sista Sista, Ohmine, Kolonia 175.3
Young Roses, Paliais, Nett 174
Kapinga Pride, Pohn Rakied, Kolonia 162.5
G-Babes, Meitik, Nett 159
Ohnonlong, Wone, Kitti 158
Beauty Cousins, Madolehnihmw 152

Second and third were only separated by 1.5 points, fifth and sixth by one point. As a statistician I have a preference for rubrics that generate more spread. I am all too keenly aware that small differentials are not statistically significant and are not likely to be repeated. That said, a dance contest is not unlike a sprinting race, crossing the line a few hundredths of a second ahead of another runner is the difference between gold and no medal.

Each group had four scores, one from each of the judges. The range from the lowest score to the highest score was smallest for Kapinga Pride and largest for G-babes. G-Babes divided and decorrelated the judges.

Dance group score spread
Dance group score spread

Beauty cousins also saw a large range in scores.

The judges seven scores, one for each dance group, tended to distribute in a range from 35 to 45.

Dance judges score distribution for the seven dance groups
Dance judges score distribution for the seven dance groups

Two judges had rather symmetric distributions about median scores of 40 and 41.5. My median of 45 was high and asymmetric. My low score of 32 was not an outlier, but that was in part due to my large inter-quartile range. One judge had a low outlier and the highest upper whisker at 49.

The dance instructors, coaches, and advisers are most likely to want to know their strengths and weaknesses against the rubric used.  Overall the category “Always smile? Ashamed? No singing with the music?” (termed “Facial” in my analysis) scored the lowest.

Criteria Average
Costume 8.46
Facial 7.82
Movement 8.14
Performance 8.48
Practiced 8.62

One of the judges last week noted that she wanted to see more eye contact. Engage the audience, smile, show confidence, and show that you are enjoying yourself. I noted that one of the dancers seemed a little stiff, reserved, and was not moving as fluidly as I knew she could. I asked her and she said she was nervous tonight. During the free style, free dance at the end of the evening, however, her smile beamed out and she threw herself into competitive dancing with nothing short of gay abandon. All of her grace and fluidity were back.

One group looked over-practiced. They did not smile, just went through the paces. Perfect synch, no life, no zest. Maybe too many hours of practicing the dance over and over. The first place winners danced with confidence, big smiles, lively. They knew their moves, but they also appeared to be having fun with their dance and seemed to relish the spotlight. They were up there to bring it, as their group name suggests, and they were clearly excited at the change to perform in front of their friends and family.

Dance criteria averages by group
Dance criteria averages by group

The G-Babes were judged to have the best costumes with a 9.75 average for the four judges. In the other categories Bring it on girls captured the top averages.

Last week I noted some correlation differences among the three judges.

Correlations Leilani Dana Kiyoshi
Leilani 1
Dana 0.748 1
Kiyoshi 0.735 0.655 1

I correlated well with Leilani, and she correlated moderately well to Kiyoshi. The items that correlated Leilani and I were not those that correlated Dr. Umezu and I, and we saw a lower correlation. This pattern occurred again tonight.

Correlations Dela Cruz Ichikawa Lee Ling Umezu
Dela Cruz 1
Ichikawa 0.19 1
Lee Ling 0.88 −0.10 1
Umezu 0.18 0.58 −0.07 1

Dela Cruz and I were highly correlated, we strongly concurred. Dela Cruz was only weakly correlated, at best, to Ichikawa and Umezu. I was not correlated to Ichikawa and Umezu, with a relationship between our scores that was no better than random. Ichikawa and Umezu were only moderately correlated.

A more detailed study of the correlations by criteria suggests that the four judges concurred on costume and performance, saw some limited agreement on movement, and disagreed on facial and practice criteria. Of interest was that the disagreements did not always occur between the same judges in those criteria where disagreement occurred.

Ultimately the rubric is open to interpretation. There was no training on the rubric nor were any of the judges professionally trained in dance. The goal was to identify groups that could be called upon to dance in dinner shows for visitors and guests. What might impress a tourist is not necessarily the same as what is likely to impress a professional dance judge. I suspect the judges achieved the goal desired despite some issues of internal inconsistency and differing interpretations.

The greatest divergence of scores seemed to be around G-Babes. I was left wondering whether there are cultural differences that may impact how one views the G-Babes.

G-Babes dancers
G-Babes dancers

The G-Babes were the youngest dancers dancing, maybe five or six years old, maybe the eldest is seven or thereabouts. I wondered about the G-Babes myself, but during the free dance sessions they were the first to stream out onto the grass and they danced and laughed with such enthusiasm. They were having the time of their lives. They clearly loved to “shake it” and enjoyed being out there with the “big girls”.

Pacific island free dance
Pacific island free dance

One cannot appreciate their diminutive stature compared to the other dancers until one sees them next to a Sista Sista dancer or a Bring It On Girls dancer as in the above photo.  Whatever differences the judges perceived, the G-Babes were clearly the crowd favorite. Ferocious amounts of cuteness, clearly well practiced dance routines, and having fun.

Sista Sista, Ohmine
Sista Sista, Ohmine

That said, what does the visiting tourist come to see at a dinner dance show? What expectations are there? All of the dances tonight were Polynesian, and only Kapinga Pride is of Polynesian heritage. The true dances of Micronesia are wonderful and awesome, but nothing like what a tourist imagines. Besides, to some extent the Micronesian dances are reserved to their cultures. They are dances with meanings and cultural import. For the Micronesian dancers, the Polynesian dances are both what the tourist expects and what is culturally more comfortable to deliver up to foreigners. That is only my opinion, but if accepted, then groups such as Bring It On Girls and Sista Sista are the dinner show dance groups.

As I noted last week, I know some of the dancers, their families, and where they from on the island. I was impressed with all of the groups. Everyone clearly had gone home and put in a lot of solid practice. The costumes were also stepped up and improved. One could see a lot of work, effort, and pride had gone into each group’s preparation. I was certainly proud of all the dancers tonight.

Kaycie Dikepa and Yolanie Lucky
Kaycie Dikepa and Yolanie Lucky, Kapinga Pride

Counting pillars and posts

The students were paired off on day one and told to count the number of pillars and posts on campus. The only definition was that the pillar or post should be standing free from a structure – a not a wall element.

Counting pillars
Counting pillars

I did not mention how to count the two-story tall pillars for the A and B buildings. This led to a natural introduction of the importance of definitions. The counting took 25 minutes, some finishing faster, singular groups slower.

M8 M9 M10
275 258 266
228 261 281
152 281 283
312 286 285
316 298 285
305 293
309 296
310 326
sample size n 5 8 8
275 unspecified how A, B bldg pillars counted
258 A, B, pillars counted as one top to bottom
305 A, B pillars counted as two top to bottom


A good sample description must include clear definitions. In the 9:00 class the students noted that they were counting the pillars differently for the A and B building. Some counted the pillar top to bottom as one pillar, some counted “per floor” and wound up with double the number of pillars. I suspect this happened at 8:00 as well.

Nancyleen
Nancyleen tallies data

Note that the 152 is an outlier: that group did not go all the way to the cafeteria – they each thought the other one had gone there. This provided a chance to note that z-scores can be used to help identify outliers that could potentially be data errors.