I've got 2 groups of 5 people. The groups differ in regard to some characteristic which I'm studying.
Data is collected in a state of rest, then both groups are subjected to a condition, after which data is collected again. Data I care about is mean pitch, standard deviation of pitch and pitch range, all in Hz.
I would like to be able to see if data is significant on the individual level, if there is a significant increase in Hz from rest to condition. Earlier studies have accomplished this through one-factor ANOVA, but I can't as I don't have enough conditions, only 2. Is this kind of statistics lost to me? It feels like it is.
The other end is just checking if results differ significantly on a group level, looking at comparing the variance between conditions and groups. What statistical method do I use for this though? Please discuss this with me, I'm lost.
>>8078816
Bamp. I need your help, you're my only hope.
Look up student's t-distribution. It was devised for working with very small sample sizes and unknown population SD.
>>8079024
doesn't let me do what I wanna tho, right?
>>8078816
Why do you think you can't do an ANOVA?
In case you have just 2 groups, use t-test. 3 or more -> anova
>>8080705
Because I want to look into single participant results on only 2 conditions
>>8081684
the thing is before I'm looking at groups I want to look at individual participants. which seems impossible
>>8082668
That doesn't make sense
You can't have significance on an individual level because significance is in terms of the population
>>8082768
I'm basing my research on a study in which it's possible to calculate one factor anova for subjects separately if you have 4x conditions and 3x variables collected for each condition, where am I wrong?
>>8082901
What's that study called?
I can't see why they would want to do that
>>8082768
Not if you have repeated measures. Then you can compare all trials from one individual in one condition to another individual's trials.
>>8083084
There's a hickup though, the sample sizes aren't even as they are data from recordings, differing slightly in length. And no, I can't just short them down to equal size.
>>8083373
link me the study you cunt
>>8083373
>And no, I can't just short them down to equal size.
Yes you can if you do it randomly, that's why they're called samples.
If you want more specific advice you'll have to provide more details of the characteristic you're trying to measure, and why you need individual-level data.
Also calling bullshit on the study >>8082901 here, unless >>8083084 holds you can't get an ANOVA for individuals since there's literally no individual-level variance to analyze. (You can get a prediction interval, though.)
>>8083373
Type 3 sums of squares for anova with uneven sample sizes
>>8083398
actually thought I replied to you at the same time as the other guy but reply countdown musta stopped it, sorry.
The study's called "Phonetic manifestations of cognitive and physical stress in trained and
untrained police officers". Middle of page 132 is where they talk about the ANOVA.
