Hi Teachers and Friends, I have a binary dependent variable (low trust/high trust) and I am performing a binary logistic regression with bootstrapping (5,000 resamples). I am trying to understand how to report the results:
Should I display the results provided by the table “Bootstrap for variable in the equation” (lower table) and add the column of the Exp(B) provided in the table “Variables in the equation” (upper table). Then should I discuss the odd ratios?
Please provide an example of how you would discuss them (both if the dependent variable is continuous and if categorial so you are sure about the interpretation too). If not, what else I could do?
Thank you so much for your kind help.
Best wishes,
Tuyen
My suggesyion is do to it as in the attachment then add the bootstrap summary as in amy other analysis
Best wishes David Booth
Thank you so much Prof David Booth! Yours is so advanced, I will try to learn from you! Best wishes, Tuyen
When reporting bootstrapped binomial regression results in SPSS, it is essential to include the following information:
Here is an example of how to report bootstrapped binomial regression results in SPSS:
"The bootstrapped binomial regression model was statistically significant (χ² = 25.35, df = 3, p < 0.001), indicating that the model provides a good fit to the data. The odds of the outcome variable (success) were significantly predicted by the predictor variables (age, gender, and education). Age was a significant predictor (β = -0.03, SE = 0.01, z = -2.70, p = 0.007), such that for every one-year increase in age, the odds of success decreased by 3%. Gender was also a significant predictor (β = -0.84, SE = 0.30, z = -2.82, p = 0.005), indicating that females had 56% lower odds of success compared to males. Education was not a significant predictor (β = 0.14, SE = 0.10, z = 1.38, p = 0.17). The pseudo R-squared value was 0.27, indicating that the model explained 27% of the variation in the outcome variable. 95% confidence intervals for the coefficients ranged from -0.06 to -0.01 for age, from -1.42 to -0.27 for gender, and from -0.05 to 0.34 for education."
Dear Assoc Prof Hammad, Thank you so much for your kind and helpful advice for interpreting and presentation of the results.
I really appreciate your kindness and will integrate your advice in my writing.
Best wishes,
Tuyen
Hello Tuyen Thanh Le. Are you asking how to report your results in a manuscript you wish to submit for publication? If so, the answer will likely depend on your discipline, but also on guidelines from the journal you have chosen. Meanwhile, you might get some ideas from this article:
https://www.scirp.org/journal/paperinformation.aspx?paperid=70962
HTH.
Hi Bruce! Thanks so much for your great advice! You offered the same paper my sup did ;) I am very happy now to have a well presented paper on bootstrapping for LR method for reference!! Best wishes, Tuyen
https://www.youtube.com/watch?v=mNrxixgwA2M&t=735s
P/S: Dear Teachers and Friends, I found this tutorial video on reporting bootstrap LINEAR regression, I wish there is also one for bootstrap binary logistic regression, in which not only B could be validated/bootstrapped, but the 95% CI for Exp(B) (odds ratios) be validated too.
I tried to validate some of the odds ratios, i.e., bootstrapping Exp(B)s, and some of the ORs are different. Thanks for your prompts for the future.
Dear Professor Ma'Mon Abu Hammad, Thank you so much again for your kind advice for me in interpreting and presenting the regression results for my phd thesis!
I would like to have your further advice and teaching on how I could work out the 'z-values' for the regression models - I was looking but could not see the z-values in SPSS (I only saw Wald statistics). I would be very grateful if you could again kindly help me to understand this and if there was a formula in case I need to manually calculate the z-value would be so grate.
Thank you very much and best wishes, Tuyen
For the Wald tests with df=1 in your output, z = SQRT(Wald value). In other words, the Wald tests you see are Chi2 tests. HTH.
Thank you2 so much Bruce Weaver !! For your valuable advice!! Best wishes, Tuyen
Hi Bruce Weaver, sorry to ask you again. I calculated the z and 99% of the variables in my model got z >.196 (in a model, I got only one relationship with z
Here's an example discussion: "In our binary logistic regression analysis with bootstrapping, we examined the relationship between trust (low/high) as the dependent variable and several predictor variables. The results from the bootstrap analysis, as shown in the lower table, indicate that the predictor variable 'age' has a coefficient of 0.25 (SE = 0.08, p < 0.01). When considering the odds ratio (Exp(B) = 1.28, 95% CI [1.10, 1.49]), we find that for every one-year increase in age, the odds of high trust occurring increase by 28%. This relationship is statistically significant (p < 0.01), suggesting that age is a significant predictor of trust."
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression