Sample Size and Trials Explanation

Sample Size and Trials Explanation


How Much Data Is Enough Data? 

● The number of items you test under a certain set of conditions is the “sample size.” 

● Each time you test a sample or group of samples, it’s called a “trial.”


As you design your test, you must plan for how much data you’ll collect. To have a successful project, you need a good-sized data set.  Depending on the type of project, you can do this either by testing many samples or by testing a few samples several times...or both. 

How much is enough? You may have heard that three data points is the minimum for an experiment. Three is okay for grade school but certainly not for older students. Why? Because more data increases the accuracy of conclusions. Many data points help cancel out unwanted influences: a one-time measurement error, a bad sample, luck, etc. So it’s important to collect as much data as you can. 

Time and Money: It would be great to have hundreds of data points, but a researcher also has to factor in how much time they have, and how much money is available for supplies. The best you can do is to try to collect enough data to have a strong argument that your conclusions aren’t thrown off by outliers. 

Your advisors can help you decide how many samples/trials are reasonable for your project. 

Trials vs Samples: How do you know if you need to focus on a lot of samples or a lot of trials?It depends on your test samples. 


● More Samples, Fewer Trials: Some samples can only be tested once because the test creates an irreversible change in the sample, so each sample is useful for only one trial. In this type of experiment, you need many samples, and test them all together in one trial. 


Example:  Grow plants from seeds and see how much they grow in a set period of time with different light conditions. You can’t “ungrow” a plant and retest it! Instead, you need to test multiple plants. So if you get 30 plants to test - 10 for each light condition - you have a respectable 30 data points to work with. 



● Fewer Samples, More Trials: If your experiment does not permanently change your samples, you may be able to test them over and over, so you can work with a smaller sample size while increasing the trials (number of times) you test each one. 

Example: Test bats to see which has better rebound, aluminum or wood, which could help a batter hit farther.   


You test this by measuring the bounce height of a baseball dropped onto aluminum versus wood bats. In this example, you only need a few bats of each type because you can test each bat multiple times.  Each time you drop the ball onto a bat, it’s a trial. If you test three bats of each type five times each, you get 30 data points. 


● One prototype, multiple trials, improve,then multiple trials again, and repeat: In engineering prototype projects, it is usually not practical to build copies of the same prototype for testing. Instead, the designer builds one prototype, tests it, makes some changes to the prototype, tests it again...and keeps doing this until the prototype meets the design requirement that was set, or time/money runs out! 

Example: You want to create a hovercraft that can travel over water to deliver packages. You set criteria of the hovercraft being able to carry 5 kg of cargo (about 11 lbs) a distance of 5 meters in 30 seconds. With your first prototype, you run a 30-second test for 10 trials. Its average distance is 2.7 meters. You make improvements and test it again, 10 trials.  This time, its average distance in 30 seconds is 3.8 meters.  You work on it again and after 10 trials, it gets a best average distance of 4.7 meters in 30 seconds. Now it’s getting close to science fair time, so you decide to stick with what you’ve got and test it another 10 times to have some extra data to show how it performs. 







In each example, the final total of data points is 30.


3 types of sample/trial groupings, same total data points:

Number of 

Number of 

trials per 

Samples 

sample

Total data 

points

More samples, less trials: plant growth after 1 month

30 (divided into

3 groups of 10)                   1 

30

Less samples, more trials: baseball bat bounciness

6 (divided into

 

2 groups of 3)                       5 

30

Prototype, 3 iterations of 

hovercraft

1                                        3 groups of 10 

30


  1. Always state the number of trials as (N = number).


By Deborah Bogard All rights reserved



    • Related Articles

    • Buckeye Science and Engineering Fair (BSEF) Sample Judging Card

    • Sample Projects

      Sample Projects Hypothesis Based Research Grade 6 Sample Project https://projectboard.world/oas/project/the-battle-of-the-shades-dwnoyz Hypothesis Based Research Grade 8 Sample Project ...

    • Science Day Teacher's Guide

      SCIENCE DAY TEACHER’S GUIDE July 2024 WHY SCIENCE DAY? Student Question: "When will I ever use this stuff, anyway?" Project Type: Student- originated, personalized, inquiry-based hypothesis-based / meta-analysis or engineering design projects. ...

    • Quad Chart Template

    • HBR/MA Research Plan Template

      Hypothesis Based Research and Meta-Analysis Research Plan **When working through this document, complete ALL Action Steps. Keep all underlined bold headings. Delete all instructions after you complete each section.** Title Name(s): Grade: School: ...