I have several predictor variables and an outcome variable, diabetes self-management which has several questions related to diet, physical activity etc. Each question has a discrete response. For eg, how many of the last seven days did you follow a diet plan? The response ranges from 1-7. But the responses are averaged to get a mean score to calculate the total diabetes self-management score.
Since the outcome data is not continuous, can I still use linear regression? I did a preliminary check on my outcome data and it is concentrated towards 7 and 0.My search of the literature suggested that the type of data is not as important as the assumption validation of the residuals. I also checked the residual assumptions of simple linear regression? But it doesn't look like it meets any assumption.
Any information on how to move forward will be highly appreciated?
The problem seems complex, your best choice would be to contact a local statistician and build with him a correct analysis after thoroughly discussing with him the experimental design and your questions.
No, you can't use an OLS regression because your dependent variable is discrete. Probably calculating a mean score is not what you want to do. Please see this link for possible possible assistance. Best wishes, David Booth
https://www.researchgate.net/post/Can_you_help_with_statistical_analysis_advice
Also as Emmanuel Curis suggests this discussion best takes place prior to data collection.
Multiple regression is an extension of simple linear regression. It is used when we want to predict the value of a variable based on the value of two or more other variables. The variable we want to predict is called the dependent variable (or sometimes, the outcome, target or criterion variable).
Amjad Karimi : that's typically one of the worst answer that can be made to the question, since it hides all the reflection about what you want to do behind an artificially apparent technical solution.
Note that you swapped the command (as.factor is for making a numeric variable a qualitative one [so you will *never* get continuous data from it], and as.double to making something into a real number [so the result has no reason to be discrete, other than the inherent discreteness of number coding in computers, just check as.double( c( "1.5", "-1.3457")]), that in general it will *not* do what you expect (especially, a discrete variable will remain discrete after as.double, even if the storage mode changes) if it does something at all, and the result may be completely meaningless (as.double on a factor, typically).
Best regards,
I feel the best you can do is convert it to a binary response such as 'satisfactory' and 'unsatisfactory'. A
above mean value as 'satisfactory' and below mean as 'unsatisfactory'
- Badges
- Science topic