Binary logistic regression using SPSS (2018)
ฝัง
- เผยแพร่เมื่อ 10 ก.ย. 2024
- This video provides a demonstration of options available through SPSS for carrying out binary logistic regression. It illustrates two available routes (through the regression module and the generalized linear models module). If you wish to download the data and follow along, you can do so by going here: drive.google.c...
For more instructional videos and other materials on various statistics topics, be sure to my webpages at the links below:
Introductory statistics:
sites.google.c...
Multivariate statistics:
sites.google.c...
![[LIVE] MCHOICE & MINT AWARDS 2024 at TRUE ICON HALL, ICONSIAM](http://i.ytimg.com/vi/wMewUeT5nbY/mqdefault.jpg)
Sir, the world needs more people like you.
That's very kind of you for saying, Mehdi. By the way, in case you are interested, I put up a new video in 2019 (there's also an accompanying dataset and Powerpoint) on logistic regression here: th-cam.com/video/cpWSSJHuT2s/w-d-xo.html
Best wishes!
DEAR VIEWERS: I have recently uploaded new (shorter) videos on binary logistic regression that cover (a) procedures for carrying out analyses in SPSS and (b) interpretation of output. Additionally, I have included links underneath the video descriptions that contain SPSS files and additional learning materials to supplement the video presentations. SO PLEASE BE SURE TO CHECK OUT THE FOLLOWING LINKS: th-cam.com/video/cpWSSJHuT2s/w-d-xo.html and th-cam.com/video/P0VKXrHrFUc/w-d-xo.html . I also have additional videos on multinomial and ordinal logistic regression, as well as mediation and moderation, that you can access more quickly through the following playlist: th-cam.com/video/cpWSSJHuT2s/w-d-xo.html&list=PLITf5JBUH0l7MD1RmMa74iyaTef0u_Moj
can you please make a video on binary probit model with interpretation of data
@@dr.nazneenarahoque8259 Hi there. My apologies, but I just saw your posting. Here is a link to a video I made awhile back: th-cam.com/video/pjK_KHsHVvY/w-d-xo.html . I hope you find it useful.
@@mikecrowson2462 thank you 🙂
DEAR VIEWERS: I wanted to let you know that this is a much older video on binary logistic regression. My newer presentations (with Powerpoints and data) can be found at the following links: th-cam.com/video/vab9NezxpBc/w-d-xo.html (2021 video); th-cam.com/video/cpWSSJHuT2s/w-d-xo.html (2019 video). I hope you consider checking them out as well!
Thank you for your videos sir
I've learnt more from you than from my tutors
Excellent description of two ways of doing logistic regression in SPSS
really helpful for running my thesis, thank you so much! Hope you may upgrade your tools to include wacom pens for writing over the screen
I have one quarry.. plz ans ... When I do analysis output not appear there was warning like" the dependent variable has more than two non missing value. For logistics reg. The dependent value must assume exactly two values on the case being processed. Execution of this command stops." ....
Now what can I do😥😥😥...
Plz ans...
I am doing a project on rainfall prediction using logistic regression in spss software 18 version... Whenever i run ,it shows a warning like "The dependent variable has more than two non missing values.For logistic regression ,the dependent value must assume exactly two values on the cases being processed.. Please help to me, how to rectify the warning
Thank you for your valuable explanation. I have question about the odd ratio. Could we consider the odd ratio that you explained in minute 20 is consider adjuste odd ratios?
Hello sir, I would like to ask to get the crude OR for binary logistic regression, I have to only put one independent variable under the covariates is it? So I have to run each IV to the DV separately?
Does we can open our paper from this FCC bundle machine??
If I want to check the standard errors from odds ratio, slope, p values. how can i check the standard Error from the existing table?
why variables in equation do not appear in block 1?
Dr Crowson,
Hi there :). I have emailed you with a log-reg output as I did not want to share that in public view, and hope you do not mind. I will continue studying your videos, they truly explain things in great fashion that is not typical considering the content. Amazing work and speaking for everyone here, we cannot thank you enough - KM
Hi. Thanks for sending (although it hasn't come through as of yet). Your topic sounds pretty interesting for sure. And thanks for the feedback. I'm glad you find the videos useful. Hang in there with the thesis!
Dr Crowson, the email address provided by You Tube was incorrect so I resent it to the correct address. Hope to connect soon and thank you for the encouragement.
23:16
what do you mean by that? I really don't know how that confidence interval is interpreted?
would make it clearer? CI is for E(B) and the all the values of E(B) are located in between the CI interval. ??
??
Thank you for your smart explanation! I have got alot lesson from it.Thank you again!
Thanks for your feedback! Glad you found the video useful!
Very nice explanation
Hey there! What kind of test should I use if I have a dichotomous DV but my IV has more than 2 groups
Thank you for the video. I have a question. In my logistic regression analysis, the Hosmer Lehman Show test does not give a p value. What does this mean? Can I continue to test?
Sir, suppose there is group of 60 consisting of 20 students, 20 professors and 20 office workers. They eat a lot and i want to find out how many of them had hamburgers (Yes/No) last night. when i form a 3*2 crosstab, I find none of the professors had hamburgers. After doing binary logistic regression, I find the value of "B" & "S.E." unusually high for the professors and the "Exp(B)" is 0.000
What is the explanation of this & how can i overcome it?
Thank you very much. very detailed and helpful
Hi Mr. Crawson, I am doing an analysis of covid acceptance (binomial yes and no)[dependent variable]
and level of education (scale), race/ethnicity [independent variable]. should I do a binary logistic regression? please advice
This is incredibly useful, thanks so much
Hi there good sir!
I did the same thing as you did, but no matter what I do, the "variables in the equation" section for block 1/step 1 does not appear.
You need to categorise the variables (when putting together DV and IVs, click on categories and put them in
I did this in my thesis but I got the odds ratio of one predictor variable in lakhs. What to do for that?
Great! I love when people do 0 for male and 1 for female.
This VD valuable for me. Thanks a lot.
I want to run a Binary Logistic Regression model but instead of looking at the predictor variables independently, I want to know how the interaction of those (categorized) predictor variables impact the Odds of a player being injured (dependent variable). How would I do that? Every video I find always just independently looks at the predictor variables but so many papers I read look at the interaction of those predictor variables on the dependent variable and give one Odds Ratio value from this
Hi Cameron, I don't know if this will help, but I do have a video on interaction effects in logistic regression. See here: m.th-cam.com/video/Dbwq_or9I0M/w-d-xo.html
It might at least give you some ideas. Cheers!
How do i get references number please ?
It is good advice for me.
I did a log reg.
I got nagelkerkes R2 = 1.000
Hosmer lemeshow also 1.000
Classification 100%
But i have 0 sig. variables
Every variable has 0.999 or 1.000
If i remove some my nagelkerkes R2 goes down to 0.4 and i get sig. variables but that shoulnd be the solution.
What can i do?
The problem is i did a linear reg before and i should so the log regression afterwards
In my linear reg. I get good sig. Variables and korr. R2 of 92%
So is there anything i did wrong?
I looked up some books and videos but still i cant figure out the problem
I hope you can help me lile your videos did but now i am stuck right now
Hi Dr. Crowson, what would you suggest someone do if they included 2 similar variables which helps increase the percentage predicted accurately but removing 1 of the variables decreases the hits drastically?
When you say they are "similar" are you meaning conceptually or empirically? Are they highly correlated and that's why you eliminated one - in an effort to eliminate a redundant variable? If the two variables are empirically distinct (even if you think they are conceptually similar) and each are contributing to the model, then it's easy to see how removal of one can alter the hit rate. That is because that rate reflects the correspondence between actual group membership and membership predicted by the model. If you eliminate a key predictor that offers unique predictive power, then your hit rate can suffer. I'd check the correlation between the predictors and ask myself whether it makes sense to view the variables as containing redundant information, necessitating the elimination of one. For example, measures of self-esteem and self-efficacy both reflect self-perceptions. And one might gather than these variables would be correlated. However, they also reflect different kinds of self-perceptions, and we would expect that measures of these perceptions would not be tapping fully redundant information. If I am predicting some certain outcome variables, I could see how eliminating one of the two might yield a decrease in overall model fit. Hope this helps.
Very well done. Thank you
How do you report the data in word? I saw this: De variabelen X1 (b=...,Wald chi2(df)=...,p
Hi, I'm not sure where you got this, but from the layout it looks like a report of the regression coefficient for a specific predictor called X1. The ... following the b= is the actual value of b (the regression coefficient from the output, whereas the ... following Wald chi^2 is the value of chi-square associated with your test of the regression coefficient. The df is actually 1 for the test of each individual regression parameter in the model. As an example taken from the video: Political interest was a significant predictor (b=.876, χ²(1)=21.859, p
Hi I'm doing quantitative research .My topic of research is Post Evaluation of Compliance to Environmental Legislation on Polythene bags ban in Islamabad: A Retailer/Vendor Perspective .
Kindly tell me which type of logistic regression should I use.If you have proper video kindly share it I;m in great problem.Please reply me as soon as possible
Top vid! thanks a ton!
Hello thanks to your tutorial is want your help since I am conducting my thesis on title factors affecting private investment performance. Therefore how could I formulate binary equation and analysis
Is it necessary to have several covariates, or is it okey to only have one?
You can have one or more predictors in your model. Cheers!
You seems to be quoting figures and terms like confidence level , significant level etc without even interpreting what does it mean to say it's high or low, or significant. Is that how you teach in class ?
Markie, please note the comment I provided below that there is a much newer (higher quality) video I put up on binary logistic regression back in summer 2019 that goes much more into detail on running analyses and interpreting results, etc. There are also supplemental learning materials available underneath the video description. I hope you consider going here: th-cam.com/video/cpWSSJHuT2s/w-d-xo.html . A second 2019 video in that series is found here: th-cam.com/video/P0VKXrHrFUc/w-d-xo.html
Hi Mike and others: How can you tell which value (within dichotomous variable; like 1=Agree or 2 = Disagree) the output is referring to? In other words, is the output based on the agree cases or the disagree cases? How do you know?
Hi Jamal, the outcome variable in this video is intention to donate, so the outcome is a binary variable coded 0=do not intend to donate, 1=intend to donate. By default the menus I demonstrated in this video will have SPSS treat the lower numeric value associated with the DV as referring to the reference/baseline group and the higher value as being the target/comparison group. So 0=intend to donate refers to the baseline category and 1=intend to donate refers to the target group. If you had a dv, where you had 1=Agree and 2=Disagree, the 1=Agree group would be treated as the baseline category and the 2=Disagree group would be the target group. If you go through the Generalized Linear Models menu, you actually have the ability to specify which group on the DV you want treated as the reference group instead of relying on the defaults associated with the route I described in this video. FYI, I have a couple of much newer - and better quality - videos (along with Powerpoints) that elaborate both routes if you. Video 1: th-cam.com/video/cpWSSJHuT2s/w-d-xo.html Video 2: th-cam.com/video/P0VKXrHrFUc/w-d-xo.html
Hope this helps. Best wishes!
very informative! thank you!
What does a unit of change in a categorical predictor variable mean exactly?
Hi there. I can't recall what exactly I said. Basically if you have a dummy variable (coded 0 and 1), then the 1 unit difference is the change from membership in group 0 to membership in group 1. So, the slope would be reflecting a group difference. Sorry, I probably wasn't as clear on that when I put this together. FYI, I have a newer video on this topic that has additional resources you can download under the video description: th-cam.com/video/cpWSSJHuT2s/w-d-xo.html Cheers!
i am happy with your tutorial, but how to validate the data for binary logistic regression model? thanks!
Dr Crowson
Do you have videos of forward and / or backward Wald, or stepwise logistic regression ?
Thank you.
Hi there. I just posted a video on this topic here: th-cam.com/video/nW8AAzBImTw/w-d-xo.html
Although I focus on the forward and backward LR tests (my preference), the Wald test approach uses the same general logic.For a bit more detail on the Wald versus LR tests you can go here: stats.idre.ucla.edu/other/mult-pkg/faq/general/faqhow-are-the-likelihood-ratio-wald-and-lagrange-multiplier-score-tests-different-andor-similar/
Hello sir, it was a good explanation, thankyou for that.. btw, how do we consider the correct percentage in the classification table as good or bad? is it has to be more than 50%? thankyou
Hello Sir, thank you for the video. It will be very helpful if you can make a video based on categorical data
what to do if the variables are not significant in last table
Hello: Non-significance is interpreted as indicating no relationship between the IV and the DV. There's not really anything else to "do", except report the IV is not a significant contributor to the model. best wishes.
I wanted to interpret what the variables DOG, EPE & CFC mean in the data you used.
Sir, It would be very helpful if you share your data, so that we can run our analysis in SPSS and match with your results.
Hi Sayanti, thanks for your posting and request. I actually have the data linked under the video, but it may not be obvious to folks since I didn't mention it in the video. Here is the link: drive.google.com/file/d/13vJ_GnjlKwCEWX7hB-CJME1Tu5jcJydQ/view
Also, this is a much older video on binary logistic regression (and not as high quality as my newer videos). I have newer presentations (with Powerpoints and data) that you can find at the following links: th-cam.com/video/vab9NezxpBc/w-d-xo.html (2021 video); th-cam.com/video/cpWSSJHuT2s/w-d-xo.html (2019 video). I hope you consider checking them out!
what if the omnimbus test insignificant? and what if the intercept also not significant?
hey did you find the solution for this?
Awesome! Thank you!
How to do collinearity test based on binary logistic regression?
Hi, unfortunately there is no official collinearity diagnostics option available in SPSS like you can get with linear regression (which would give you the VIF or Tolerance). That said, you can screen for highly correlated predictors by examining a correlation matrix of the predictors, where problems would be present with r's in the .80s or greater. The downside of that approach is that it doesn't take into account for the possibility that one predictor is a linear function of one or more of the others. The variance inflation factor (VIF) is computed as 1/(1-R-square), where R-square is the regression of each predictor onto the remaining predictors. So although the program doesn't give you the VIF, you can compute it pretty simply using this formula [note, you are computing VIF per predictor]. Conventionally, values > 10 indicate more substantial problems with collinearity. You can read more here: www.statisticshowto.datasciencecentral.com/variance-inflation-factor/
@@mikecrowson2462 Thanks for your immediate resposne. So, I wonder what can I do in this siutaion regarding to the issues of collinearity?
@@zhengjiaren6548 Are you asking if you find problems with collinearity what should you do? If that's the case, you have various options. One option is to remove variables that appear collinear with the others from the regression equation. Another option might be to create a composite index from the collinear variables (kind of like summing or averaging items from a survey). Another option (if you have a number of predictor variables) is to perform some type of data reduction (e.g., PCA;see e.g, th-cam.com/video/6Ycw3_iQHD0/w-d-xo.html) and reduce the larger set of variables down to a smaller set for entry into your model. In the end, these are possibilities but your decision should be guided by your research goals (e.g., theory testing versus prediction only). Hope this helps.
Very useful. Thanks.
Thanks for watching!
Can someone advise on what to do if multiple binary categorical predictor variables ? Confused on how to or what variable to control for etc
For my thesis submission in 2 days :(
Hi there. There is no inherent problem with adding in multiple binary categorical predictors (you're essentially treating them as dummy variables). You're just adding them in the regression just like you would continuous predictors. The question of "what to control for", however, is a conceptual/theoretical one and not a statistical one per se. Control variables are generally selected for inclusion in the model based on the researcher's consideration of other possible variables that may "explain" a portion (or even most) of the relationship between a focal predictor and the criterion variable. For example, researchers often use "gender" as a control variable when the possibility exists that gender shares variation with their predictors and criterion measures. In this case, they simply wish to "partial out" gender from the relationship between other predictors and their criterion measure, in order to reduce potential bias in their regression slopes. Of course, there is a trade-off when adding in multiple control variables, as the model becomes increasingly complex and collinearity among predictors may occur (due to a large number of predictors). So, you do need to be "choosy" when deciding on covariates to include - and again, this is a conceptual/theoretical issue. My suggestion is to run your question by your thesis chair and discuss what covariates she thinks ought to be included as controls. Just my two cents. Hope this helps.
Dear Mike, Thank you so much for the reply. It made a little bit more sense. However my question perhaps better explained if I inform you of the study:
DV = Belief in the Human Soul (Yes / No) , numerous IVs , all categorical, most are binary (1/2) or three levels (1, 2, 3). Examining childhood religion and beliefs, belief in after life etc as an adult, whether or not science can be used to study the human soul and standard SES variables.
The control I used in the past was "Did you grow up in a household that was part of a religion" because most responses were yes when offered yes or no
If doing hierarchical log-reg, or a binary log-reg, how would I do Enter, and / or step wise? I hope I make sense and cannot thank you enough
Ok, I understand. You CAN do a hierarchical version of binary logistic regression, and (wouldn't you know it) I have a video that touches on this: th-cam.com/video/dW-99z_XTLs/w-d-xo.html
You'll have to use a little syntax in SPSS to do the hierarchical logistic regression. I will say that I do worry about whether you will have enough variation on your covariate to make it worth including it. Nevertheless, you can give it a try and see if it works. Hope this helps.
Thanks
thank you for elaborated information!
what can I do if all independent variables were not significant?
Basically, this means that the model isn't correct and that you need new predictors, to put it simply
please i need certain video on how to interprete the output result of binary logistic regression
Hi Keb, click on the links I provided in the video description. You will go to videos that provide a substantial amount of information that you are looking for on interpreting results. Thanks for visiting!
why couldn't my lecturer explain it this simply ?
I partially understood SPSS before I watched this video. U do not explain what certain things do and how they play any significance in the overall analysis.
Hi Alex, this was an older video of mine. You might find my more recent videos helpful. You can also download pointpoints under the video description for more of a 'blow-by-blow' of the results and what they mean.
My most recent video is here: th-cam.com/video/vab9NezxpBc/w-d-xo.html
An older video from 2019 is here: th-cam.com/video/cpWSSJHuT2s/w-d-xo.html
I hope you check them out and find them useful.
-best
This is such a great explanation! However i do have one critique, it would be really nice if you could work on avoiding the use of "uhm". you use it every other sentence.
Ini naratornya bukan orang inggris asli ya? Bahasa inggrisnya kacau banget wkwk jadi bingung ngedengerinnya banyak kata sambung yg ditabrak2:')
Tapi videonya bermanfaat banget kok,
Tetap berkarya pak
Try this video. It's newer. th-cam.com/video/cpWSSJHuT2s/w-d-xo.html