I want to investigate the ability of a group of students to learn programming concepts in one hour.

The students are grouped by ages: 6 - 17. The tasks are divided into concept and type. Each concept has example tasks, to correct code, to copy code and to write code without help. I measure the number of tasks completed and the time elapsed in each one.

What is the best way to analyze this data? And to represent them?

UPDATE

I wanted to compare the performance of students by age and gender. enter image description here

I draw a simple bar chart with the means of the tasks performed correctly.

But the experiments have different durations and much dispersion. There is some more reliable statistical method than analyzing only the mean.

I am not an expert in the subject and neither in statistics.

  • 4
    Do you have a hypothesis (if so, include it in the question), or will you solely do post-hoc analysis? Or, in other words, what is your research question? – Steven Jeuris Aug 10 at 10:36
  • 1
    I'm mildly confused what type of answer you're expecting. Would you be okay with references to papers showing similar studies/experiments? Or are you only looking for a step-by-step guide? I ask, because I want to make sure you get the information you're looking for. – Seanny123 Aug 11 at 16:55
  • Seanny123, It is okay references to papers showing similar studies. I want to show that the 'Hour of Code' can be useful to learn programming concepts. – Marina Aug 11 at 21:59
  • 1
    I'm not sure this is really on topic for psychology - seems like the question is more about pedagogy? Certainly ed psych is part of psych but I don't see much of an ed psych bent here. Others may certainly disagree with me, though. Might be better for cseducators.stackexchange.com though I'm not personally familiar with that community. – Bryan Krause Aug 11 at 23:34
  • 1
    As Steven asked; what is the hypothesis? Are you using two conditions and do you wish to test the effectiveness of study method A versus B? Or is it a descriptive study? Do you expect an improvement over time or something? The number of completed tasks is a vague descriptor. Do you mean the nr. of correctly accomplished tasks? – AliceD Aug 13 at 13:09
up vote 1 down vote accepted

If you want to report the dispersion, add some error bars, or better do a box plot, or even violin plot, but the latter require understanding of less introductory concepts (for tuning purposes). For a non-graphical presentation of e.g. age by task (and by gender as well), see contingency table. These can be nested.

If you want to know if there are significant differences between the groups, read about the F-test. If want to test diffrences between individual tasks, you can do that too (e.g. t-test), but you have to worry about adjusting for multiple comparisons.

I will not go into further details because this board is unsuitable for giving you a stats 101 education. And so is Wikipedia, by the way; its articles on such topics are generally too technical for the mere practitioner of statistics. It's best you talk to someone who has had at least that stat 101 level of education, although it would be better with someone who at least [still] practices statistics on some level. All else failing, read an intro textbook of applied statistics; or see coursera for a free class etc.

You should also learn how to use a software package that does all of these, e.g. R or SPSS. Depending how good you are at programming you may find one or the other easier. With software like these, you can also easily do plot arrays that basically iterate over a covariate's (e.g. gender) level (i.e. values). There's no standard name for these, R calls them facets; SPSS calls them panel plots.

Your Answer

 

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.