Hello everyone,
I am looking for statistic advise.
I have a small sample (86 participants) of which 30 are husband and wife (15 couples) and the rest are either single or have a partner but the partner is not in the sample. The 85 are all from a specific category (not clinical but similarly specific, hence the small sample), not just a random community sample. My study looks at measures where there it is possible that a husband and wife might be more correlated to each other than to other study participants across some of my dependent and independent variables.
1. How would I go about testing whether the couples are more correlated with their partners than other study participants (both the other couple and non-couple participants)?
2. Supposing that my testing in (1) shows that couples are more similar to one another, how could I incorporate this when looking at my analyses (my currently planned analyses are: correlation coefficient, multinominal regression, and Hayes' Process Mediation Modeling)?
Thank you for your time and attention.
I don't well understood your topic, but I think You can deal with Cluster Analysis , MAnova test. for correlation between couples [x_(i),y_(i)], i: 1,2,...,15. a Paired data reformulation is required to check the difference between the couples... Yes as you mentionned you can use multinomial analysis with a random effect, but the sample size in each sub-group is critical for estimation accuracy.
Hello July,
Here's a suggestion that would be pretty straightforward. Consider your couples set (30 cases) as one batch of 15 sets. Then, randomly pair the remaining 56 cases into a second batch of 28 sets. For each paired set (couple or random), compute the correlation across the measures of interest. You'll now have 15 correlation coefficients in batch A and 28 in batch B.
The comparison of interest is, how often does a randomly selected correlation from batch A exceed that of a randomly selected correlation from batch B? (There are 15 x 28 = 420 such comparisons.) If the proportion is well beyond .50 (which would indicate no difference), then you have evidence of higher correlations from the matched couples than from randomly matched cases. This is analogous to the Common Language Effect Size statistic of McGraw & Wong, determined from an exact enumeration, rather than an approximation assuming some specific distribution of scores.
Norman Cliff modified this to a dominance statistic, in which the result is p(batch A > batch B) - p(batch B > batch A), so that 0 would indicate no difference.
A more complete test would be an exact test, wherein the 15 matched case correlations were compared with the complete distribution of correlations arising from all possible pairings of the 86 cases (=3,655 pairings possible, including the 15 matched case pairings). If the 15 matched case pairings consistently show up among the top percentile points of that distribution, this confirms your hypothesis.
Good luck with your work!
References:
Cliff, N. (1996). Ordinal methods for behavioral data analysis. Mahwah, NJ: Lawrence Erlbaum. pp. 124-140.
McGraw, K. O., & Wong, Sp. P. (1992). A common language effect size statistic. Psychological Bulletin, 111, 361-365.
Dear July Lies ,
Correlation is a concept that refers to two variables.
If two variables are such that they are dependent on each other, then the two variables are called correlated variables. If the dependence between them is linear, the correlation can be called linear correlation. Otherwise, the correlation is non-linear.
The linear correlation is measured by Kearl Pearson's coefficient of correlation while the non-linear correlation is measured by correlation ratio.
In your study, the variables are not specified/defined.
You have a small sample (86 participants) of which 30 are husband and wife (15 couples) and the rest are either single or have a partner but the partner is not in the sample. But, what about the variables between which you want to know the correlation. Also, where are your data for determining correlation ?
Correlation may be between the attachment of husband and wife, sharing of responsibilities between husband and wife etc.
You need to describe the variables and to collect data on the variables.
Only then you will be able to proceed further.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression