Descriptive abstract

Download 33.5 Kb.
Size33.5 Kb.
Data Documentation Files

NAME: SeniorSwimTimes.txt
TYPE: Three years (2009, 2011, 2013), cross-sectional (565 individuals).
SIZE: 565 observations, 20 variables

Data consist of 500-yard freestyle swim times for male and female swimmers age 50-94 in a biennial national competition. Variables include year, gender, age, age group, swim time, seed time (qualifying time from state competition), and split times (in each 50-yard segment).

2009 times provided by FastLane Tek, Inc. (; 2011 and 2013 times provided by Hy-Tek Sports Software (

Data are tab delimited and missing values are coded as *.

Obs: observation in sorted list (year, gender, age group, place)

Place: finish order in age group for that year
Name: participant name (omitted)
Gender: 0= male, 1 = female
Age: age in years (50, 51, …, 94
Age2: age squared (for non-linearity tests)
AgeGrp: age group (1=50-54, 2=55-59, …, 9=90-94)
Year: competition year (2009, 2011, 2013)
Seed: qualifying time prior to national competition

Time: time in national competition

Diff: difference between seed and time
Split: time (seconds) in each 50 yard segment (Split-1, Split-2, … , Split-10)

There are 5 missing seed times. In 2009, one extreme outlier (male) was omitted (assumed data error). In 2013, one extreme outlier (female) was omitted, and one male whose time was clearly incorrect. Two other males with incorrect times in 2013 were included because sufficient data existed to calculate their times. In a few cases where two 50 splits were missing but 100 split was available, the 100 time was averaged between the 50s (orange highlight). Inconsistent split times or those that could not be estimated (more than 2 missing) are shown as *. Data are from


Data are from an athletic event (500-yard freestyle swim) from three biennial U.S. National Senior Games Association (NSGA) summer competitions (2009, 2011, 2013). The Summer National Senior Games are held during odd years (ex: 2009, 2011, 2013, etc.) To participate in the National Games, you must first qualify through an NSGA State Games during the even years (ex: 2008, 2010, 2012, etc.) Participants must be at least 50 years old during the qualifying year. You may qualify through the state you live in, or any State which allows out-of-state competitors. In most sports, the top 4 finishers in each age group qualify for Nationals. Data collection (“munging”) involved tedious re-formatting, cross-checking, editing to spot inconsistencies, and combining several data sources. The resulting “cleaned” data set is easily understood by any student, and can be used for a variety of tasks in basic statistics classes. The author was a participant in the 2011 NSGA competition (Houston, Texas) as a member of the OPC Swim Team (Michigan).

Suggested tasks for students: Calculate descriptive statistics (center, spread, shape) for Time and Seed. Compare descriptive statistics for Time by Gender and AgeGrp (describe and interpret observed patterns in plain language, and suggest possible reasons). Correlate Time with Seed. Perform simple regression of Time on Seed (does national competition help athletes achieve better times, on average?). Calculate descriptive statistics by Split (Split-1, Split-2, …, Split 10) (describe how swim time changes as the race progresses, and suggest possible reasons, e.g., strategies employed by the athletes). Plot split times in various ways (e.g., by Gender and AgeGrp). For each gender separately, perform multiple regression of Time versus Age and Age2 (to look for non-linearity). Repeat using Gender as a binary (students could test regression coefficients for equality). Consider interaction variable Age*Gender. Does Year have a significant effect on Time (e.g., creating binary variables for Year)? If so, suggest possible reasons. Should a separate analysis be done for each year, or should year be included as a binary? Advanced students could try other ways to model the effect of Age on Time (e.g., using AgeGrp binaries, or quantile regression (e.g., 25%, 50%, 75%) for each gender separately. Does Year matter (as a binary)? In each analysis, students should look for high-leverage observations and unusual observations. If there are outliers, what should be done (if anything)?

See SOURCES above.

Name: David P. Doane
Affiliation: Oakland University
Surface address: Department of Decision and Information Sciences, Oakland University, Rochester, Michigan 48309.

e-mail address:

Share with your friends:

The database is protected by copyright © 2019
send message

    Main page