Adjusting for confounding - Learn all about adjusting for confounders in this SPSS tutorial
HTML-код
- Опубликовано: 21 авг 2024
- In this epidemiology tutorial, I will teach you how you can deal with confounders in epidemiological data, including stratification and multivariate analyses. In addition, I will show you with real data how you can do this in SPSS.
The data that are used in this video were previously published here: link.springer....
About me:
I am a registered clinical epidemiologist and working as a fellow in medical oncology in the Netherlands. I have published over 50 manuscripts and have received several large research grants. I am particularly interested in research in geriatric oncology and am a an active member of the International Society for Geriatric Oncology.
You can find more information about my work on my linkedin page:
www.linkedin.com/in/nienke-de-glas
And here is my full bibliography:
pubmed.ncbi.nl...
Disclaimer:
Views and opinions are my own. Examples from clinical research will always include either my own work, or previously published research.
Hey I love how you teach us these little tips and tricks such as "paste" and where to click if we want to quickly go back to the previous windows to configure the model! Love from Malaysia.
thanks a lot for your nice words, highly appreciated :)!
Your videos are must-haves for PhD students! Thank you so much! :)
Thanks I learned so much from you about statistics and research than I ever did in faculty. Stephan from Africa.
R
Thanks a lot, that is great to hear!
Hi Dr. Nienke. I am from Indonesia. Thank you for your great explanation, it is really help me to solve my confuse to analyze with adjusting variables using logistic regression model for my research article.
Thanks, Dr. Nienke, this lecture is more useful. God bless you
Thanks a lot, that is great to hear!
Thanks so much DR de Glas. It's an amazing video with a thorough explanation.
Very easy to follow and understand. Great video
Super helpful videos. Have watched dozens of others, but couldn't get it. Very practical and very easy to understand
Thank you!!
Thank you for your videos. I'm learning a lot. I also enjoy watching this because of your beauty.
Your tutorials are really great and straight forward, and super useful!
One of the best simplified
Thank you so much!
Thank you so much Dr de Glas for an excellent video. It was incredibly helpful. Bless you, for your contribution to science and education.
Thank you very much. Your explanation is very helpful.
Well explained! Thank you
This video helped a lot. Thank you very much, Dr Nienke.
You are welcome, thank you for watching!
Good explanation. I need more of your tutorials
thank you very much for the compliment! do you have specific topics you would like to see?
Great. Thank you, from Ethiopia
That was very clear, you helped me a lot. Thanks
Thank you so much, you saved my day.
Thank you Dr. Nienke, this video was very simplistic and useful. I have a question, what should I do if I have multiple covariates which (possibly interact with each other) and my study wants to see independent association with each. Should I consider them confounders for each other?
Thank you
Thanks, Doc. This was really beneficial.
So useful, thank you!!! 🙏🏼😊
Great explanation thank you
You are welcome and thank you for the compliment!
I agree.
Thank you so much.
Thank you.
This is really great. Can you present a lecture on analysis of longitudinal data?
So helpful thanks
Thank you for your nice lecture. I am kindy looking at some explanation on overmatching and introduction of confounding bias to a study. Thank you!
Thank you for your compliment! I will put it on my list of future videos!
Very good
Thank you for uploading this video. very comprehensive & useful. may I request to upload the detailed method to calculate adjusted Risk Ratio in SPSS/?
I am about to do calculate results for my dissertation and I think if I watch this about 70 times I will have a vague understanding of what to do 🙈 Thank you ❤ do you do tutoring?
That is what's nice about video recording; can pause and replay until saturation is achieved :)
Hi doctor. I found a statistically significant relationship between Depression scale scores and disease recurrence in a cross sectional study. There's no relationship between age and depression scores in the study. Can I assume age is not a confounder in the relationship between depression scores and recurrence rates?
In statistical analysis, determining whether a variable is a confounder requires considering multiple factors, including statistical significance, prior knowledge, and study design, as was mentioned in the video. While the lack of a relationship between age and depression scores in your study is an important finding, it does not automatically imply that age is not a confounder in the relationship between depression scores and disease recurrence. Here are a few points to consider:
1. Statistical significance: If there is no statistically significant relationship between age and depression scores in your study, it suggests that the association between these variables is not strong within the observed sample. However, it does not guarantee the absence of a relationship in the broader population or that age cannot act as a confounder.
2. Prior knowledge: It's essential to consider prior knowledge and existing literature on the topic, as Dr. Nienke de Glas stated. Age is a widely recognized and relevant factor in studies related to mental health and disease outcomes. Therefore, even if your study does not find a relationship, it is advisable to examine the existing body of evidence on the relationship between age and depression scores.
3. Study design: The cross-sectional design of your study provides a snapshot of data collected at a specific point in time. It may not capture the temporal relationship between depression scores, age, and disease recurrence. Longitudinal or cohort studies with appropriate follow-up periods can help establish causality and identify potential confounders more effectively.
To determine whether age is a confounder in the relationship between depression scores and disease recurrence, it is recommended to conduct further analyses or review additional literature to assess the consistency of findings across different studies. Consulting with a statistician or an expert in the field can provide valuable insights into the specific nuances of your study and guide you in making informed conclusions.
@@garristotle thank you for this detailed response. I'll do more literature review upon your guidance.
Thank you for the explanation. One question, if I may: I noticed that when you chose age and comorbidity as covariates, the p value for comorbidity was not significant. But when you chose age, comorbidity, and polypharmacy as covariates, the p values for all 3 were significant. What does this mean about the status of comorbidity and whether it has an impact on surgical complications by itself or whether it is not significant?
This most likely means that there is a statistical interaction between comorbidity and polyfarmacy, which makes sense as patients with multiple comorbidities also use more medications. We did actually test for this and indeed there was a statistical interaction, it was just a bit too much for this video.
For the interpretation: do not focus too much on this, it is much more important to decide which (clinically relevant) confounders you think are out there and use these in your model, rather than focusing on the statistical interactions.
@@nienkedeglas_mdphd Thanks a lot.
@@RawaMuhsin you're welcome!
Great video. Would you adjust for confounders in the same way if you did cox regression?
Yes indeed, the same principles apply there!
Could you please help me understand how to perform a ROC analyses in SPSS using age as the test and complications as the state, BUT adjusting for the cofounders (morbidity and polipharm)? Thanks
Thank you very much for your excellent presentation and if possible i need to know how to demonstrate deterministic and probabilistic bias analysis by using SPSS or STATA
Hello Dr de Glas, thank you so much for your very clear video. I am trying to run an analysis on a dependent variable that changes over a span of 2 years while I analyse the effect of two independent variables on the dependent variable. (depression at baseline is 100%, changes in depression after 2 years) I think i need to correct the dependent variable at baseline before I start my analysis but I don't know how. HELP please!!!!
Thanks so much this has been one of the most helpful videos.
I have a question. When I use dummy (0-1) variables to include variables with multiple levels into a MV analysis, SPSS sometimes excludes one of the variables automatically. How should we address or interpret this?
Also sometimes before adjusting the model for certain variables I noted the OR was less than 1 and after the MV analysis OR for same variable is > than 1 . Is that possible , how should I interpret?
how can I adjust non-categorical data before applying t test? for example I would like to, again with logistic regression?
Nice one
Thanks!!
Hello Dr. de Glas, thank you for your great explanation of this topic. If I want to adjust for confounding in linear regression or GEE model, how should I operate and interpret the result from SPSS?
@just somebody Thank you!
completely agree :).
Thanks a lot！ please can we use this hierarchical regression model with categorical dependent, independent and control variables? or is it possible to it with generalized linear model method for regression?
how we are doing adjusted cox reg. analysis
hi Dr. De glas im doing case-controls study im not sure how will i do AOR.
queen
are you able to make adjusted KM curves? Also, how do you adjust for continuous variables?
unfortunately, this is not possible, since KM only allows for univariate (unadjusted) analyses. You should instead use a cox regression model in which you can enter all variables, it does not matter whether these are categorical or continuous.
you can also watch my video on survival analyses, it may help you to understand it better!
thank you for this great video ! just one question- what does the sig means in every variable value? for instance in age(1) the sig is .688 (above 00.5)
thank you for your question! this is the p-value, so you would report this as 0.688. Since this is above the (generally accepted) 0.05 cutoff, this would mean that the factor is not statistically significant in your model.
@@nienkedeglas_mdphd I really appreciate your response!
The other values of the "age" variable, however, are significant. How does this impact your model? Do you need all the values of this categorical variable to be significant in order to say that the variable as a whole is significant?
@@omeravraham7343 Good question! I usually tend to report (and interpret) the overall p-value that is presented at the reference category. This is the overall p-value that tests whether there is a trend among all categories (so increasing age = increasing OR).
interesting .how to analyses wealth index from asset? thankyou.
what are impact of adjustment in relation to relative risk
Good question! THat depends on the interaction between the confounder, the determinant and the outcome. If you would adjust for a factor that is actually not a (strong) confounder, the impact is much smaller compared to adjusting for a strong confounder.
How can I contact you I need help in identifying confounder in my research on epidemiology
Congratulations @Nienke de Glas ma'am for your publication in Breast Cancer Res Treat (2013) 138:561-569. Can we get the data set to practice
Thank you
Thank you
Thank you