I would like to understand how much of a DV is explained by a number of IVs.
My data set is relatively small = approx 300 in total. The data was collected at three time points so its longitudinal (approx 100 at each time point). Furthermore - my data is quasi experimental so i am using the same intervention and control group at each time point.
So back to the question.
1) Can i do 1 multiple linear regression with all my data as one set (all 300 repondents)?
2) Do i need to do three tests - one for the data at each time point /baseline, T1 and T2)
What option is the most correct?
You could do a multiple linear regression with a quantified DV as follows:
DV = a + b1 (treatment/control (1/0)) + b2 (time-ordinal or actual) + b3-bn (covariates).
this is an analysis of covariance. there will possibly be interactions and collinearities to explore
Thank you James Leigh .
I would like to understand how much of the variance a number of variables explain the outcome variable employee driven innovation (EDI).
In this analysis i am not so interested in if the sample is from the intervention or control group or which time point the data was collected. Can i treat the data set as a whole or should i do the analysis separately for each time point and group in case time or the intervention impacts how much of the variance in EDI is explained by the different variables i measure?
If you are not interested in time or treatment you can pool the data and Rsq will give proportion of variance "explained" by the varables measured.
If you do wish to consider time and/or treatment the analysis of variance of regression will allow for these and let you estimate the proportion of variance explained by the measured variables.
It's impossible to answer this question, I think, without knowing more details about your data and the relationships you expect!
Linear regression is never "correct" or incorrect. It makes you see what you want to see.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping