I have a unidimensional marital satisfaction scale (Johnson, Zabriskie, & Hill, 2006) with five items. My sample was 194 married individuals, and using the data from my sample, I wanted to re-examine the reliability. The Cronbach's alpha was .53 which was well below than the typical alpha I would get from this scale.
I tried to observe the data distribution from each item by using the box plots, and I got this results. I do not know how to interpret the box plots and the IQR for item 1 and item 2.
My question is: how will you analyse or interpret the participants' responses on this scale?
Thank you for your kind help.
Dear Edwin Adrianta Surijah ,
it is unclear what value you are plotting on your box plots. A much better approach would be a principal component analysis which would provide you with a weighting for each item on your first component. You could then remove any item with a weighting below 0.4 and use the regression scores to create your scale rather than simply adding items together. Please see my study guides for more information.
The SWML-scale items seem to consist of a 7-step rating scale. I assume you are using jitter (random noise) to separate the responses (dots), but not for computing the boxplots? So, then, the dots between 5.5 and 6.5 are actually the scale reponse 6 (agree) while for the boxplot they are all sixes. You can then conclude from the plot that the reponses are highly concentrated at the scale response number six for items 1 and 2.
The plot seems to explain the low alpha coefficient estimate for the scale: there is little variation in the items and hence there can be little covariation or correlations between the items. Alpha is essentially based on the average item correlation / covariance, and when these are small, their average will be low and alpha will take a low value.
I think that you have used the ggplot or ggplot2 package in R for these plots, so I would encourage you to download the psych-package and use the alpha-command for exploring the relationships in the data. You can obtain additional information about the scale such as inter-item correlations, which may be helpful.
The bottom line interpretation might be that your subjects are rather satisfied with their marriages.
Best regards,
Timo
Timo Törmäkangas thank you for taking your time to explain this to me. You were right with the jitter to separate the responses. I used the ggplot2 package to obtain these plots. I also used command alpha from the psych package to get the alpha values. You were also correct with the low adjusted item-total correlation. Based on your explanation, I would interpret these plots as: Most items were positively skewed, however, item 1 and item 2 had lower variability compared to other items. On item 1 and item 2, most participants' responses were concentrated on the scale of 6.
Again, thank you for the kind help.
yours,
Edwin
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics