I have a repeated measures design, looking at how sex hormones effect number of responses on a dichotic listening task, three times over the course of the menstrual cycle.
Where it simply the number of correct responses given by the left ear (LE) or right ear (RE) I wouldn't be stuck, however I'm also looking at voice-onset-time (VOT).
Each question is two dichotically presented consonant-vowel syllables (one LE, the other RE), and each pair is made of a combination of long VOT syllable (delayed onset of the vocal chord vibrations after an unvoiced consonant such as p/ t/ k) and a short VOT (immediate voicing: b/ d/ g) (also short-short and long-long, for a total of 30 combinations - after excluding b+b/d+d etc combinations).
So, the answers I receive are separated out into a wealth of data such as number of correct responses in the left ear from a short-long combination (LE-SL, meaning the short VOT syllable was presented to the left ear, and long to the right) and likewise the number of correct RE responses to the same VOT combination. Or, number of correct responses in the right ear from a short-short combination (RE-SS) etc. (As an aside, I will also have the number a VOT combinations with an incorrect response where neither syllable was correct).
I can then total these aspects to give total number of short VOT-left ear responses etc. So I could now compare means for left-ear-short and left-ear-long, or left-ear-short with right-ear-short.
However, I then have to introduce time into these analyses and I get a bit muddled. Will I have to compare each totalled-response variable separately over the three phases (IV with 3 levels) and so use a repeated measures ANOVA, or is there a way to compare the differences between say short-VOT and long-VOT over three time periods? (attempting to see if a previously demonstrated long-VOT advantage is stable over the course of the menstrual cycle).
This is in SPSS by the way.
Please help, I have bamboozled myself!
Many thanks to any who may offer assistance. (attached is what the collation of responses will look like)
It seems to (if I understand your design correctly) thet you could use a 3-way repeated measures ANOVA on the %% correct answers of the participants as dependent variable. You have 3 within (repeated measures) factors:
- Sessions: 3 levels
-VOT: 2 levels
-Ear: 2 levels
Inspect the histograms of the %%. If they look fairly normal, go ahead. (Remember that the Shapiro test is biased by size. You (only) have 32 participants , so this test should not be significant. On the other hand, the ANOVA is known to be quite robust against (a.o.) non-normality of ditributions. Also check the sphericity test. If significant, use the corrected p-values.)
Thanks so much Jos Feys
Sadly, I may have fewer than 32 participants, as the study is yet to begin. I was merely attempting to structure the way in which I would capture the data and run analyses.
Would the VOT not be 4 levels however? (SS, SL, LS, LL).
I also realised last night that I can calculate a laterality index for LE-RE (positive number showing right or negative showing left ear advantage) and similar values for all combinations of variables, reducing (SSxLE, SSxRE, LExLE, etc etc) down to +/-0 values for each state (so a -66.67 value for LS demonstrates a strong left-ear advantage).
With this could I assess LE/RE index (1 DV Mean per phase) over Menstrual Phase (IV, 3) to measure the degree to which it changes over time? Similarly, I would have four measures of VOT-laterality (+50.00 for LL, showing a dominance of the right ear, or -25.00 for LS etc). Would this cause any problems?
I should admit I am quite confused about what you would like to do. As I understand, you would like to find out whether sex hormones affect consonant-vowel syllable accuracy. To measure this accuracy, you would simultaneously present pairs of monosyllables to your subjects, one in the Left and the other in the Right ear. Each pair is a combination of Short or Long VOT (voice-onset-time), with 4 possible combinations: SS, SL, LS, and LL. Your dependent variable is the number (or proportion) of Left Ear (LE) correct answers, e.g. presented in the left ear from a short-long combination (LE-SL, meaning the short VOT syllable was presented to the left ear, and long to the right). In this example the subject correctly identified the monosyllable with a Short VOT. Am I wright? In this example, you particularly want to compare these number of correct responses to a short VOT monosyllable, when it is presented in the Right Ear (RE-LS combination). You are interested in this instance in the Left – Right Ear distinction. This corresponds to the first repeated measures factor: Ear, with 2 levels.
Secondly, you are also interested in the VOT combinations: SS, SL, LS, and LL. In my opinion this corresponds to the second repeated measures factor: VOT combination, with (indeed) 4 levels.
Thirdly, you also are interested in the menstrual cycle, which 3 Sessions (repeated measures).
I understand that you are interested in differences in accuracy between Left and Right Ear, but I would not calculate difference scores. You can assess these differences with testing the effect of Ear.
So, the 3-way repeated measures ANOVA on the %% correct responses – in my opinion – still holds.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS