Comparing results for computer and pen and paper writing I am conducting a study to examine the differences between writing paragraphs in English among junior high school students in Japan. 
We collected writing samples from four classes, approximately 38 students per class, and looked at the number of words written, the number of grammar errors, and the number of spelling errors. 
The students wrote five paragraphs over the semester, with two classes writing three computer paragraphs and two paper and pen paragraphs while the other two classes wrote two computer paragraphs and three paper and pen paragraphs. 
We assumed that students would write more and make fewer errors using a computer because of the spell and grammar checker. We will calculate descriptive statistics, but I am wondering how we can show if there is a difference between the computer and paper paragraphs. 
I am thinking that we will use a t-test for independent samples comparing the means of all computer writings with all paper writings. Is this correct? 
I have only used a t-test with two distinct groups (male and female) but in this case it is the the difference between two activities. Any help would be appreciated. I am a real novice to statistics. Thanks
 A: This is a very complex problem for a novice. An independent t-test (and other relatively simple methods) assume that the data are independent. A matched t-test only allows data that are in pairs.
Your data are not independent and not in pairs. In fact, they are dependent in a couple different ways: You have repeated data on students and you have students nested within classes. That is, what student #1 wrote by hand is related to what student #1 wrote on computer; what student #1 wrote at time 2 is related to what he/she wrote at time 1, and what a student in class A wrote is more related to other students in class A than to students in class B.
In addition to that, your dependent variables are not continuous, but counts.
All this suggests that the correct model is a nonlinear multilevel model. This isn't something that novices would know how to use (no offense intended - you indicated you know little statistics). Perhaps an analogy would be asking a student in his/her first semester of English to translate Shakespeare into Japanese.
So, you can go with descriptive statistics (even that is a little challenging here - you will want to break them down various ways), but if you want to model the data, I suggest finding an expert. 
