# How To Reject Null When Unequal Samples? Should I Use Welch's, ANOVA or What?

#### Wendy12

##### New Member

Hello,

I am in the 7th grade, working on a science fair project. I engineered an electronically enhanced vision device, which I call "XYZ glasses." I have built a prototype and tested it with 17 different people. I am now summing up my work, and I have graphs that make it seem that my glasses do work better than no glasses. However, I would like to analyze my data with a statistical test. But, what test do I use? Can someone please help me? I will do the work, just tell me what test I should research and learn how to use.

What test do I use to reject the null in this example? My t hypothesis is : Do people see better with XYZ glasses? The null is therefore that people see equally well without XYZ glasses.

17 subjects were tested as follows: First, as a control a person was tested without XYZ glasses, could the subject read 3 words in fog at 20 feet, with a score of 1 for each words read. Thus if Mr. X read all three words his score was 3, if he read no words his score was 0, etc. Second, the same person wore XYZ glasses and was tested just like before, but with different words. The subject was also tested 4 times while wearing XYZ glasses. So, Mr. X could have 4 scores for tests, e.g., 3, 3, 3, 2.

This was done with 17 different people, so I have a sample of 17 "control" scores, and then for the same people, I have a sample of 68 (17 people times 4 tests each) test scores.

I will learn how to do the calculations, but I am in the 7th grade, I am currently taking algebra 1, so I do not know how to figure out what test to use to compare my data and to test the null hypothesis. I think this is "paired" since the same population was both the control and test group, and it is unequal samples of 17 and 68. What test do I use? Welch's T test? Student's T test? An ANOVA? Exact Fit? Please help. I do not know, can you please help me?

Thank you

Last edited: