I'm trying to measure the level of significance of devoiced vowels in different linguistic contexts; whether or not certain contexts cause vowels to be devoiced.

Basically my data right now is a table with the contexts in the horizontal line (4 of them), and speaker on the vertical (speakers are not a variable, I just keep it for identification). The values filled in are yes or no (I left "no" as blanks), meaning yes this context caused devoicing , or no (no value) this context doesn't devoice this vowel.

How can I calculate level of significance of my data? If the result was actually significant due to different contexts or not.

This is the last part of my results chapter but I'm having trouble figuring it out.

  • $\begingroup$ Was each context tested with different speakers? Than that would be a contigency table and a Chi-square-test of independence. I guess however, that the same speakers were examined in all 4 contexts, thus resulting in dependent samples. See en.wikipedia.org/wiki/Cochran%27s_Q_test whether Cochran Q test fulfills your requirements (let "context" be "treatment" and "block" be "speaker"). $\endgroup$ – Bernhard Nov 9 '16 at 13:06
  • $\begingroup$ no the same speakers for tested four times. im reading about Cochran Q test and it seems right, thank you. $\endgroup$ – Luna md Nov 9 '16 at 13:36

If your observations are independend, i. e. each speaker in only one condition, than you can count your observations and build a contingency table like this (assuming, it's all about the moon)

          full   waning   new   waxing
voiced  |  15  |   22   |  10 |   13
devoiced|  10  |   12   |  11 |    7

A Chi-Square-Test of independence could answer the question, whether voiced/unvoiced has anything to do with the moon phase. The Q-Test I mentioned in the comments is for dependend samples, Chi-Square is vor independend samples.


Your Answer

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

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