Science Buddies: "Ask an Expert"

My topic is "eco friendly household detergents vs store bought detergents"

The problem with my project is that I cannot eliminate the factor of human error in my experimental procedure.

The human error is that:
1. when I am testing the detergents I use a spray-bottle, and there is a possibility of spraying with a different pressure each time. This affects the amount of liquid that comes out, as well as the 'cleaning power' of the liquid as pressure helps to clean (eg. Pressure hose cleans a car better than an average garden hose)

2. And also I use paper towels to wipe away dirt (I make sure to only wipe once with one paper towel piece as I use the amount of paper towels used as a measure of how good/bad the detergent works), but the problem is that there is a possibility that each time I might wipe with a different amount of pressure therefore my results could be wrong.

Someone has suggested to me that I should use a window wiper motor that I can get at a scrap yard, and attach it to a board and cover the rod in a material
- but I'm really not a handy person, so I don't think I'll be able to construct the device correctly
-I don't know what surface I'd use during the procedure(the surface which I need to make dirty and test the detergents on) as a window wiper is curve-shaped
-I'm not sure if I'll have to waterproof the motor so that it doesn't get damaged(is it already waterproofed...because it's connected to something that wipes water)
-I don't know how I will power this device (car battery? Alternating current? Will this end up being dangerous?)

Hi Seemeen,

It's good that you recognize the possibility of human error. This is a source of bias and error in essentially all experiments, but I don't think you need to try to come up with a fancy wiping machine to solve this problem. In a lab there are two things we commonly do to control for these types of problems: 1) replicates, and 2) blinding.

For 1) replicates, you would want to clean more than one area with each type of cleaner. Then, if you happened to press a little harder one time than another, it would tend to even out across all of your repeats. The absolute minimum number of replicates you should do for each condition is 3, but if you have sufficient materials to do even more, all the better. The more replicates you do, the less your results can be due to random error. Doing multiple replicates like this also allows you to do statistical testing (probably a t-test), which will give you a result called a p-value. This value gives you the chance that your results could have occurred by random chance, and we generally consider something to be "significant' if it could have happened by chance less than 5% of the time. When you get to this phase of your experiment, if you want to do some statistical testing, just post here again, and someone can help you if you don't know how to do this.

The other thing you should do is blind yourself (or whoever you have doing the cleaning) to what type of cleaner you are using. This helps control for your own bias. If you subconsciously think that the eco-friendly detergents will clean better, you may unintentionally wipe harder/more with those detergents, so you should have someone put the detergents into bottles marked only with 1, 2, 3, etc. Then, after you have your results, you can check which detergent is which for analyzing your data.

You're doing a really great job thinking through sources of error in your experiment, and I think you have an interesting and worthwhile experiment set up. Feel free to post again in this thread if you have any more questions.

JMP

Hi Seemeen,
JMP has given you some really good advice by suggesting that you do multiple replicates of each type of detergent and that you blind yourself to the type when you do the experiment. Another important advantage to having several replicates is that you will be able to actually calculate the variability in the results for each type of detergent. That information will be even more important than just showing the p-value to include in your final report!

Be sure to let us know how your project is proceeding!

PS! As JMP pointed out, you can't eliminate human error, but by having replicates, you can quantify the magniture of the error!
Good luck!

Thanks so much.
Sorry, but I forgot to mention that I presented my project at expo last year and the problem was the human error.

I'll try to follow your advice.

Thanks so much for the expert advice. I think it will really help.