0
$\begingroup$

I have a sample size of 51 and for my each participant, I have 15 dichotomous variables for vocabulary knowledge (know-not know) and 15 variable for their score for each word. Something like this:

     table     table score
P1   0              41
P2   1              55
P3   0              37  
P4   1              52

I have 15 different pairs of variables like this and I restructured the data into long format (variables to cases). So now I have 2 main variables as "words" and "scores". I will go for pearson correlation but this time "N" increases to thousands. I have 51 participants, not thousands of people. Is that a problem or normal? I think it is quite reasonable but I am not sure.

Improve this question
$\endgroup$
7
  • 1
    $\begingroup$ Exactly what will you be computing the correlation for? What do you hope it will reveal? $\endgroup$
    – whuber ♦
    Commented Nov 29, 2014 at 15:08
  • $\begingroup$ I have 15 different words. My participants have taken a test on these words if they know their meaning or not. Then they involved in an eye teacking session. I wonder if there is a relationship between knowing a word and how long it is fixated in miliseconds. Each word is dihotomous variable, how long it is fixated is continuous. I want to execute a single correlation test, not 15 different bivariation. $\endgroup$
    – Emrah Dolgunsöz
    Commented Nov 29, 2014 at 16:59
  • $\begingroup$ That is an example of a regression problem. Pearson correlation will tell you little of any value. A very great amount concerning this question has been posted on our site (thousands of posts). Take a look at this tightly narrowed search for some ideas and then broaden the search if you don't find the answers you seek. $\endgroup$
    – whuber ♦
    Commented Nov 29, 2014 at 17:53
  • $\begingroup$ I scrutinized the topics yet I could not find an exact answer. It seems like a linear regression, however I have 15 continuous outcomes and 15 dichotomous predictors. Then do I have to conduct 15 different regressions for each variable? Rather than this, I need a single regression and a single analysis for all these variables. $\endgroup$
    – Emrah Dolgunsöz
    Commented Nov 30, 2014 at 14:19
  • 1
    $\begingroup$ In your earlier comments it sounds like you wish to explore 15 relationships, one for each word. This could be performed as 15 separate analyses. However, if you believe there exists a common underlying relationship, then you would conduct a single regression analysis between test results and eye-fixation durations (in which the specific word might enter as a covariate). This would be much more appropriate, powerful, and useful than computing a bunch of Pearson correlation coefficients. $\endgroup$
    – whuber ♦
    Commented Nov 30, 2014 at 16:51

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.