I would like to run regression using training scores and DTI parameters for my participants , the problem is that the number of practical sessions and DTI scan are not equal !!
What I need to do ?
I appreciate every little help
Thanks
Manal
In my opinion 'yes', you can use regression analysis by choosing the maximum number of pairs of dependent and independent variable possible for your data-set.
The problem here is "missing data," since regression requires scores on every variable in the equation. But if the amount of missing data (non-overlapping observations) is not large, there are ways around this -- such as "imputation" of the missing data.
Hell, Manal ,
It is necessary to choose a variable with a minimum number of observations. We denote it by T. If T> 3n, then we can try to construct a regression. . It is important that T be more than n. The more, the better (if the parameters of the model can be considered constant for the sample under consideration). In the literature there are recommendations that T be more than 5 to 7 times. In practice, T = 2n is also encountered. If the model has a small error and is adequate, low data redundancy will be acceptable.
I want to add that to build a regression, it is necessary that all variables have a number of observations T.
Use the variable with the least observation as the basis for sample size so that all variables have the same observation length. This will get rid of the issue of missing data. Use only observable data.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running