1. ## power analysis

i'm helping out with the power analysis for a study that is using 6 independent variables and one dependent variable. orginially, i suggested a regression model because it sounded like the individual was looking to see what the best predictor of her dependent variable would be. unfortuantely, regression analysis usually calls for a large sample size for adequate power which is not plausible for the study. i was wondering if another statistical test would provide adequate power with a smaller sample size. perhaps, factor analysis or correlation? thank you for your assistance.

2. It boils down to how they define "best predictor" and whatever statistic they intend to use to establish the significance of the prediction.

Could you provide additional background - what are the predictor variables and what are they trying to predict?

3. Thank you for replying so quickly. The study is looking at 6 variables (i.e., dialect rating, rate and fluency of speech, vowel errors, consonant errors, word stress and intonation errors) and how those variables relate to speech naturalness. They are unsure of what statistical test to use. Thanks again

4. Are the independent variables categorical or continuous? Is the dependent variable categorical or continuous?

5. the independent variables are continuous and the dependent variable is a continuous rating scale. . .

6. Then I'm afraid that regression may be the only viable option. What is the sample size?

Maybe if this person could demonstrate the lack of power, then it may provide impetus for a larger follow-on study.

7. that's what i'm afraid of. i personally like regression and know very little about factor analysis. the individual was looking at running a sample size of about 18-25 because of the limited availability of the population of interest (i.e., native Arabic speakers who speak English as their second language). Would it be possible to run an ANOVA using native English speakers as a type of control group???

8. You could include such a control, but the overall analysis would still need to be regression since you would just be adding one categorical variable - the rest are still continuous.

Factor analysis requires huge sample sizes to be reliable, and it is more of an exploratory method rather than a modeling method or one to use when looking for significant relationships.

#### Posting Permissions

• You may not post new threads
• You may not post replies
• You may not post attachments
• You may not edit your posts