Thank you! I've struggled with SPSS and stats forever but you've made it so straightforward. Excellent explanations and not over complicating everything.
I am so grateful for your videos! I've been using them for the whole module and when i graduate in September, I owe it partly to you!!! You don't know how much this means to be and I promise once the mist clears; I WILL come back and see how I can show my appreciation 🤑Thank you David (Mr Robinson!) you are an amazing teacher, clear, concise, and effective!
Thank you very much for your good predisposition! Thanks to your recent explanations I was able to finish the results section of my research! Greetings from Cordoba Capital, Argentina.
I can also say that I am so grateful for your videos. I watched the linear regression one and now this one. Thank you so much! If anyone asks me for help on these topics, I will definitely be pointing them in the direction of your videos. Some of the best I have seen on statistics.
Hi David, I am an exercise scientist, and I deal with research methodology and statistics a lot. I have to say that this is by far the best explanation I have seen on youtube. Simple, precise, not overcomplicating. Well done!
@@DavidRobinsonPhD Hi again! Is it ok to put more than one categorical variable (binary and more than two categories variables) with few continues for multicolinearity assumption? For example: two continues and 6 categorical...and what about this assumption when I have binary logistic with only dichotomous variables (chi-square?)? thx mate!
@@gorgonije Hi Goran, thanks for your questions. I’ve looked at a few sources, and all suggest it’s fine to have more than one categorical predictor and that these can have more than two categories. Additionally, it’s common to have multiple continuous predictor variables. Unfortunately, few sources cover multicollinearity among categorical predictor variables, though, yes, some suggest that you could use chi-squared tests.
I greatly benefited from the APA writing style. There are not many resources for getting an example of how to describe the results in APA format. Thanks a lot!
This is the most helpful video I have seen on this statistical test. Clear instructions and goes through every step. I found reporting the statistics so much easier, using this video !
wow, this was so useful and thorough, i feel like i finally understand what's going on with my thesis analyses. i feel like i owe you, friend. thanks for doing this and much love to you!
Thank you so much! I watched a plenty of videos but i couldn't grasp the concept! This helped a lot! I have been stuck for days due to this! I can finally proceed now! Can you please make a video on moderation analysis?
Hi Dr Robinson - this was fantastic. Thank you for your level of detail. Wondering if you have/will do one of these for moderated logistic regression? I'm writing a thesis and have found limited number of videos on this topic?
One of the best videos on assumption checks, thank you. I have a question, what if the p value is significant but the CI range includes 1? Normally, if the CI range includes 1, p value is consistent with it and is not significant right?
Hi, May I double confirm with you about the hosmer and lemeshow test, however the significant is 0.506 (non-significant), it still indicates a good model fit in data right ?
Thanks for your videos David, they have been very useful. How would you interpret a dichotomous predictor variable such as gender (male, female) as it would not make sense to say "for every one-unit increase in gender..."
Hi James, thanks for your question. You could just use the odds ratio for the gender variable. For example, if it was about 15, you could say that females, compared to males, were about 15 times more like to have anxiety (assuming males were used as the reference group).
This was so helpful, thank you! I take it I can use this method if I have more than two independent variables? As long as I have a binary dependent variable?
Thank you! Question: why do we have to do the separate linear regression to look at multicollinearity? Why isn't it available in the logistic regression?
Hello David, If I am not wrong, you present the most complete binary logistic regression method on SPSS from inception to completion including an APA style report. Congrats, I wanted to ask if we had statistically significant p for Sex and OR 5.22 then in this case, how could we report our odds ration in our results section? Thank you in advance.
Hi Stavros, thanks for your question. It would look something like this: “The sex odds ratio of 5.22 suggests that females were 5.22 times more likely than males to have an anxiety diagnosis”. This is because I coded males and females as 0 and 1, respectively.
I am carrying out analysis for willingness to pay for crop insurance. I encountered with one problem that my Omnious test value is above 0.05 but according to the Hosmer & Lemeshow test , the data fits into the model. Can I move forward with my analysis ?
Does anyone know what to report or what is still relevant when your mdoel is not significant? right now i did it like this: The logistic regression model contained partner strength as dependent variable and SDO as independent variable. The model was not statistically significant X2(1, N = 102) = 0.46, p = .498, indicating that the model was not able to distinguish between the choice for either the stronger or weaker player. The model only explained between 0.5% (Cox and Snell R square) and 1.1% (Nagelkerke R squared) of the variability in the choice of the stronger or weaker player. Results revealed that SDO was not a significant predictor of partner strength W(1) = .456, p = .500. Consequently H3 was not supported. ANd is the Walt mandatory or could i delete that?
Thank you! Very useful! I have one question. In a binary logistic regression, are we allowed to say “explains x % of the variance”? I recall that in my statistics classes we could only say something about the model fit, rather than the variance.
Hi Rutger, thanks for your question. The example in the video is partially based on Julie Pallant’s example in the SPSS Survival Manual, in which she essentially says, “The model explained X% of the variance in the dependent variable.” Other authors use this phrasing too, such as creators of this site: statistics.laerd.com/spss-tutorials/binomial-logistic-regression-using-spss-statistics.php
Great video. I have one question, what about the assumption of liniarity of independant variables and log odds? How do we know that that assumption is being met here?
Hi Monika, thanks for your question. I didn't mention this assumption as it's not always listed in textbooks, though other videos have covered it, such as this one: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-sciPFNcYqi8.html&ab_channel=MikeCrowson
Thank you so much David for such an amazing and comprehensive video! You really saved my life! I have a question though. You said in 12:08 that "if we see a positive value here this actually means that there's an increased likelihood of having an anxiety diagnosis if the individual is female." Doesn't that entail the opposite? I mean that if the individual is male, he is less likely to have an anxiety diagnostic? If yes, can that be reported even though it is not as direct as the one mentioned? Thank you again !
@@DavidRobinsonPhD Thank you for confirming! I have another question, can I modify the process in SPSS so that I can get a specific statistic number for this interpretation? I mean instead of only making implications about men through the statistics we got for women. Thanks in advance for your help!
@@noufy9747 You could consider recoding the variable. E.g., if you originally coded the sex variable as 0 = males and 1 = females, you could make it 0 = females and 1 = males.
Hello David! Thanks for the great and informative session. Highly appreciated. Regarding the correlation assumption, do you mind sharing your reference(s) for the 0.7 cut-offs? Same with the standard residual value of > 2.5 as the basis for removing a case. Thanks again!
Hi! Your explanations are straightforward. I am struggling with an assumption my professor wants us to check. Independence of the residuals. I have one categorical variable to input with the other continuous variables in SPSS. I found that the scatterplot was not random but tightly cohesed on two lines and my histogram was not a bell curve. There was not any collinearity. Should I have not entered the categorical variable when I ran SPSS residuals?
Is it necessary to build a AUC with the results of your logistic regression model? I saw that I could use the predictive probability saved during the analysis as a new variable, then I can used it to build a AUC curve and pick a specific cut off value based on the sensitivity and specificity.
Hi Sir, thank you very much for your tutorials, very helpful! I have a question: I am currently doing my masters thesis but i struggle with making a nice results table like you did, is it possible for you to upload a template for the results table so i can adjust it to my own results? Or could you make a video on how to create these tables in word? A baseline characteristics table template would help me a lot as well. Thank you!
Hi Hector, thanks for your message. You can download APA table templates and can find guidance on APA tables here: www.everyoneneedsaneditor.com/post/apa-7th-edition-quick-guide-and-table-templates
Hi David! thank you so much for this session, it was very clarifying. I have a question regarding the classification table, I am trying to work on a model but just realized that I have a skewed dataset so my model will tend to predict more 0 than 1, I have tried to reduce my sample to balance my cases but it's not working out. Is there any correction test I can use to enhance my results? thank you!
Hi! this is so helpful thank you! what would I report/how would you write in APA style if results were not significant? would you still include all of that information?
Hi good day I would like to ask about the interpretation of logistic regression in SPSS, because I did background LR logistic regression and the analysis result provide me some choices on Step 1a, 2a, 3a,4a. Hence, I’m wondering that can I take each step of the value that I want ? Or I only can take the value is in Step 1a while I cannot take the value in step 2a
Hello Sir David, thank you for this video! May I clarify the steps you did in checking the multicollinearity assumption. You used the tolerance and VIF values in determining if the assumption is satisfied or not. Is it okay to based it according to VIF values even if one or more independent variable is not continuous (in this case, sex is nominal)?
This was amazing, thank you so much David your video is a lifesaver. I wanted to ask you whether we need to report specificity, sensitivity, accuracy of the model as well. My second question is, what would be the interpretation of if the constant OR is big as 4861670.382. Thank you in advance !!!!
David! S.O.S I am wondering if you do one on one zoom sessions? Im a thesis student needing help! Wondering if you offer this service and how to maybe contact you? :D
