I want to confirm that, is there any specific acceptance criteria for beta values as well under Unstandardized Coefficients panel of Coefficients table in linear Regression analysis. Because I am getting the B values ( 4.30E-04 , 7.79E-04, 4.15E-04, -7.62E-04, -8.18E-04 ) respectively for five Independent Variables and I am confused about the acceptability of these value.
Sabiha,
I don't know what "acceptable" means.
I presume that these are the unstandardized weights that an OLS regression rendered; they are what they are, given your choice to use this mathematical model. I am just guessing here at what the problem may be, but if the small size of these weights is confusing, note that this is influenced by the scale (SDs) of the predictor and criterion as well as the partial slope between predictor and criterion. To see the partial slope independent of these SDs, you can take a look at the standardized values. You may also be interested in the inferential test regarding each of these weights, which can be construed as a test of the partial slope, or equivalently a test of the increment to predicted variance in the criterion contributed by that variable to the remaining variables. Do keep in mind that "to the remaining variables" changes for each variable considered thusly.
HTH, but I am not quite sure what the question is.
John
Do you obtain reasonable predicted y values? If so, then your units sound fine. :-)
I agree with James. As long as the predicted y values are in the acceptable range of your outcome variable, the coefficients should be fine. There is no specific cutoff point to use to determine an acceptable beta coefficient. Assess the fit of your model using the available criteria depending on what model you fitted. If the model is fitting your data poorly, then consider other alternative models.
This is going to sound terribly rude. But your question suggests that you do not know what you are doing.
The un-standardized 'slope' coefficients gives an estimate of the change in the response in its measured metric) for a unit change in a specific predictor variable ( in terms of its measured metric), statistically holding constant the other predictors. Given that, do your results make sense? See what happens if you divide a predictor by a 1000 and refit. And now do it by dividing the original predictor by a 1000.
So what is to be done? I suggest that you go and talk to someone in your institution who does understand and/or follow a good online course: http://www.bristol.ac.uk/cmm/learning/online-course/
and test yourself on the exercises.
In general I and many others are not fans of standardized betas
https://www.researchgate.net/post/Whilst_running_multiple_linear_regression_what_is_the_use_for_stardardised_beta_coefficients
- Badges
- Science topic
- Similar topics
- Counseling Psychology
- Confusion