I don't see anything wrong. The R2 indicates the magnitude of relationship between the set of predictors in the regression and the outcome variable. The standardized beta indicates the relative contribution of each predictor to the regression equation. Looking at your correlations, you can see that all predictors are related to your outcome variable. Each predictor is correlated with one or more other predictors, so in the regression, only one predictor (symptoms) is significant. The other predictors fail to add significantly to prediction above symptoms. Had the predictors all been uncorrelated with one another, likely they would each predict unique variance and be significant in the regression. Keep in mind that your sample size is only 85, so there isn't a lot of power to detect incremental variance contributed by the other predictors. Had your sample size been larger, it is possible (although not guaranteed) that more predictors would be significant.

Dear Paul,

Thank you so much for your kind attention and answer to my questions. I calculated sample size based on G power with the medium effect size (.15), power .80, alpha at .05.  Also, literature suggested at least 20 subjects per variable. So, I chose 85 sample size. Did you see the assumptions? Is there good homoscedasticity? If so, I would be even happier. 

I got more confidence in my findings now and can present confidently. Thank you so much again.

Sincerely

Saraswati Bhandari

Faculty of Nursing

Mahidol University

Dear Michael,

Thank you so much for the detailed information. The tolerance is (Age = .844, educational level, .498, overall symptom severity = .753 and social support = .685) and VIF is age = 1.185, education = 2.007, overall symptom severity = 1.328 and social support = 1.461).  All are within normal range (lessthan 5).

Anyway, I got new insight into the results of regression analysis. Thank you so much again.

Dear Michael,

I regressed all independent variables on another independent variable. The r square was large for all. The sample size calculated based on the effect size of the previous similar study (even chosen medium effect size although that study showed 29% variance on HRQoL which is large). What will be the best option now? 

Many thanks.

Dear Michael,

Thank you very much.

Dear Michael,

I choose to do nothing and report the results as you suggested. My thesis defense is tomorrow. I worry that what to tell about the lack of precision due to high standard error (SE of regression coefficient). I was totally unaware before I asked a question here. Do you have any idea about such evidence that even with high standard error, we can report the results safely doing nothing?  Sorry for many concerns. 

Dear Michael,

Thank you so much

Excellent inputs Michael, as well as excellent advises. I agree that Saraswati mention the limitations clearly. In one week towards the defense increasing the sample is a hardship. For these researchers I would start with a sample based on 95% certainty, 5% researcher error, and choosing P=% of breast cancer ladies, q = with no breast cancer.

So, based on this info:"

Around the World: Nearly 1.7 million new breast cancer cases were diagnosed in 2012. Breast cancer is the second most common cancer in women and men worldwide. In 2012, it represented about 12 percent of all new cancer cases and 25 percent of all cancers in women"

Z = 1.96; p = 0.12 [12%], q = 0.88; e = 5%,

N = (Z)(Z)(p)(q)/(e)(e)

then N = 162. to carry out the research. If you are asked what would be a minimum number to start the research if you want to do changes.