1
$\begingroup$

I convert data from long-form to wide-form for my thesis. It makes sense intuitively for me to do so as I can easily see the relationship between every variable. However I was wondering is there a research paper out there that supports such a conversion?

$\endgroup$
5
  • 3
    $\begingroup$ The physical form (really, the format) of the data is a matter of communicating it with your software. It has absolutely nothing to do with statistical theory. There is a theory of relational database structures that includes various "normal forms"--but it is based on considerations of database operations such as creating, adding to, and updating the database rather than performing statistical analyses of the data. $\endgroup$ – whuber Apr 21 '17 at 14:07
  • 2
    $\begingroup$ This depends entirely on how the data arrive and what you want to do and what software you are using. For what I do and what I use, conversions to long from wide are about 10 times more common than the opposite, but that is not a matter of principle, just empirical. You might as well as ask which way an envelope should be handled. $\endgroup$ – Nick Cox Apr 21 '17 at 14:08
  • $\begingroup$ Thank you for your valuable comments, I generally know what both of you mentioned. However for my master thesis I was hoping to get a citation. I thought it might be better to say I did this because acording to paper X doing so is better in a similar scenario. $\endgroup$ – ljourney Apr 21 '17 at 14:25
  • $\begingroup$ That'd be unnecessary. Analysts reshape their data based on the requirement of the test and software (e.g. paired t-tests mostly prefer wide data, and independent t-tests prefer long; mixed effects model would prefer long data; and GEE tend to favor wide.) If your data structure allows you to carry out the test properly then that itself is the reason. And if by leaving the data as is, the same tests can also be done, then your restructuring is redundant even it's widely practiced. And yet, does that change your analysis results? Probably not. And if that's the case, why cite? $\endgroup$ – Penguin_Knight Apr 21 '17 at 14:57
  • 1
    $\begingroup$ You don't need a citation for something that is just a matter of convention or convenience. $\endgroup$ – Nick Cox Apr 21 '17 at 15:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.