Hello,
My weakness is statistics and I could use some advice.
My research proposal is comparing 1 group sample of people, and testing pelvic range of motion in the sagittal, frontal and transverse plane using 3D gait analysis. Each participant in the group will be tested with 3 different conditions (different types of cast).
After reading it seems as the data will be quantitative, a 'two sample special' and comparing before and after (pelvic angles whilst barefoot, then with cast, then with another type of cast) - would a paired t-test be appropriate?
Reasons why I am confused is reading t-test they talk about hypothesis (type 1 errors), but I was told you either have a research question, or a hypothesis not both. I went for a research question. I thought it may be a paired t-test as another source mentioned this measures the difference, good for before and after data (for example subject 1 barefoot compared to subject 1 with cast compared to subject 1 with 2nd cast).
It may also be important to note that because each participant will be observed in all 3 conditions there is no randomisation, which seems to be important when determining which test.
This is very hard to explain in text but would a paired t-test be appropriate?
thank you very much.
If you want a test to generalize a conclusion from your sample to a more general population, your sample should be a random sample from that population you want to generalize to (so here should be the randomization). If you took a random sample of sports students from your university that are willing to participate in such exercises, you can then generalize to the entire population of sports students from your university that are willing to participate in such exercises (further generalization is not warranted by your data, but it may be arguable with common sense).
So your data is quantitative, a measured value (don't know: degrees? cm? something like that). The next question is is it is reasonable to assume a conditional normal distribution of such values? If you can answer this with yes (and the actual data does not scream at you that this would be a stupid assumption), you can go with a repeated-measures ANOVA (paired-sample t-test is for two groups, but you have 3; the rmANOVA is a generalization of the pst-test).
Because you will apply the three treatments to each person, you will not have separate samples to compare, so this eliminates the suggestion to use either ANOVA or OLS regression.
Instead, I suggest you use a series of "paired comparison" t-Tests (T1 vs T2, T1 vs T3, and T2 vs T3).
As Jochen Wilhelm said, the selection of participants should be randomized...but also, the order that the selected participants are exposed to the conditions should be randomized, too. And, if practicable, the order of condition exposure within each participant should be randomized.
- Badges
- Science topic
- Similar topics
- Psychology