I have three parameters, 1. Attachment level with phone (categorical data eg. Highly attached, detached, highly detached etc) 2. Hours spent using phone before covid (continuous) 3. Hours spent using phone after/during covid (continuous). Now what or how to correlate these three parameters to draw a meaningful conclusion?
Mainak Ghosh If you use "correlations" you can only relate variables "two by two", or even relate two variables "controlling" by a third (see https://statistics.laerd.com/spss-tutorials/partial-correlation-using-spss-statistics .php)
I do not know the objective of your analysis or the hypotheses you wish to test, but with the information you provide I suggest that you analyze the possibility of testing a mediation or moderation model. For examples about mediation and moderation you can see:
https://www.sciencedirect.com/science/article/abs/pii/S0005796716301887
Article Mediation and moderation analyses: exploring the complex pat...
https://bit.ly/2EO1dIQ
https://www.youtube.com/watch?v=RqkGMqDU20Q
If you use SPSS in your data analysis, I recommend using the PROCESS macro that allows you to estimate models with mediating / moderating variables very easily (http://www.processmacro.org/index.html). If you work in the R environment, the "mediation" package will be very useful
https://cran.r-project.org/web/packages/mediation/vignettes/mediation.pdf
Good luck!
Dear Mainak Ghosh ,
I suggest you start with a scatter plot of after v. before for each attachment level. If the distribution on each plot is approximately "cigar shaped" you can run a Pearson correlation. Otherwise use a Spearman correlation. If the best fit lines are approximately parallel you can try to use an ANCOVA. If they are not parallel you can try a multiple linear regression with attachment levels as dummy variables. Please see my study guides for more information.
Hello Mainak,
I endorse Peter Samuels ' suggestion of first running scatterplots of your variable pairs. For the two continuous variables, ordinary Pearson correlation will likely suffice (presuming some apparent linear form to the scatterplot). For the ordinal variable (attachment level) with either of the other two, Spearman correlations would be suitable. If all you really want is relationships among the variable pairs, that's all you'll need to do. If you were trying to answer some deeper question (e.g., does attachment and prior use predict post-covid use of phone?) then some sort of regression model might be appropriate.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Developmental Psychology
- Attachment