I used this when I first learning how to create Bland-Altman plots in RStudio. Little did I know I would be subscribed to this channel years later for your psycho analyses without realising you are the same person!
Hi, thanks for the nice video. We have this problem: what happens if at the very beginning the mean difference is signifacantly different from 0? Is there a way to go further?
These aren't "confidence intervals" They are known as "Limits of Agreement" and because they are calculated with sample means and SDs they have confidence intervals of their own.
Hi, this is a great video, I really like how you go into some aspects of interpreting the plot. I was wondering, how would the mean and standard deviation be calculated if there were repeated measurements of the 2 depression scales on the same individuals?
Great tutorial but would invite some more context especially in the part about proportional bias, how is this interpreted and why do you do a 1 sample t-test of the difference.
I constructed a BA plot despite finding a statistically significant difference in the one sample t-test. I believe that there was systematic error in the two measurements (MRI vs Ultrasound) because the images had to be transformed. Therefore, I adjusted the test value for the t-test and the analysis was not statistically significant. Furthermore, no proportional bias was present. My question is: is this is acceptable? Thanks for a great video!
Hey Todd, maybe its a stupid question, but if I want to compare two methods that use different scales (both likert, but one from 0 to 2, and one from 0 to 4) then I need to standardize the scales in z before I use the Bland-Altman Plot, right?
Thanks for the info =). However, one question: Is there any restriction with respect non-continous variables? Could this method be used to compare interval measures? Thanks!
Please can I ask why you would not move forward with a Bland - Altman Plot if p value is statistically significant (i.e. 0.003)? What does a p value of 0.003 in this case mean? - Thank you, I know you made this video a while ago but it (any many others I've watched) have been very useful in my studies.
What if the dependant variables are categorical data (for example scores 1 to 9)? Can we still perform one sample t-test to see if there’s significant difference between test 1 and test 2?
Dear Dr. Tood, I was criticized by a refree for using one sample t-test before bland altman to make sure there was no statistically significant differences. So I have been looking for any literature to support my doing. I would be happy if you give me any advice.
Hello. Thank you for such an informative video. I don't if my plot is too maginified or what, only 4 dots can be seen on my graph despite there being several values. Regards.
Thank you very much for this video. However, I have my lower confidence interval falls below the limit of the plot and there is an SPSS message telling me that my lower confidence value is below the allowed limit. What could be the problem Dr. Grande please? I really started to get confused about how to solve this issue. Looking forward to hearing a reply from you ASAP please. Dr. Faraidoon Rawf
Might be slightly late... but you can double click to bring up the properties tab and go along to scale. From there you can enter the range of values for the Y axis.
hello dr todd thank for this vedio . i have learned a lots ,, i have a question sir ,, can we do bland-altman plot with 3 variables line .. and how we can do it ? thanks a lots
+Ed Marquez Confidence intervals are calculated using the SE for an unknown mean and an unknown standard deviation. The calculation in this video was based on an unknown mean and a known standard deviation, so the SD was used.
Actually... it's a sample SD and an estimate of sigma, the population standard deviation. Confidence intervals for sigma can be calculated from chi square distribution.
Thanks for the video. How do I get my chart to display the "0" line in the centre? the chart is set with the ranges of -5 to 15 and my (95%CI) ranges from LB: -8.08 to 11.00. So when I add the y-axis reference line for the lower limit...I get an error message saying it is beyond the limits of the chart!
![Bland-Altman Plot [Simply explained]](/img/1.gif)
I used this when I first learning how to create Bland-Altman plots in RStudio. Little did I know I would be subscribed to this channel years later for your psycho analyses without realising you are the same person!
God bless you Dr.Todd Grande
Hi, thanks for the nice video. We have this problem: what happens if at the very beginning the mean difference is signifacantly different from 0? Is there a way to go further?
What would be the next step after assuming a proportional bias?
Why would you not go forward with the plot if the t-test shows a result
Thank you, you are my saviour😭
These aren't "confidence intervals" They are known as "Limits of Agreement" and because they are calculated with sample means and SDs they have confidence intervals of their own.
I think so too.
I agree
Hi Todd, This is a very informative video. Thanks a lot
Loved the video, thank you for uploading it, very clear instructions.
You are quite welcome - thanks for watching -
Todd, I'm running 4 Bland Altman plots, have a B of .118 (close to 0.0) and non significant at Sig = .828, so no proportional bias. Thanks again
If the initial independent samples t-test is significant, does that mean the two measures are definitely not in agreement?
Todd, very helpful. Thank you
Hi, this is a great video, I really like how you go into some aspects of interpreting the plot. I was wondering, how would the mean and standard deviation be calculated if there were repeated measurements of the 2 depression scales on the same individuals?
Thank you Dr Grande. Very informative and simple
I'm glad you found the video useful. Thanks for watching.
Great tutorial but would invite some more context especially in the part about proportional bias, how is this interpreted and why do you do a 1 sample t-test of the difference.
I constructed a BA plot despite finding a statistically significant difference in the one sample t-test. I believe that there was systematic error in the two measurements (MRI vs Ultrasound) because the images had to be transformed. Therefore, I adjusted the test value for the t-test and the analysis was not statistically significant. Furthermore, no proportional bias was present. My question is: is this is acceptable?
Thanks for a great video!
love your videos! Only wish the voice volume was a little higher
Very helpful. Thank you!
Thank you
Hey Todd, maybe its a stupid question, but if I want to compare two methods that use different scales (both likert, but one from 0 to 2, and one from 0 to 4) then I need to standardize the scales in z before I use the Bland-Altman Plot, right?
And, does it make any sense at all to construct a Bland-Altman-Plot if the correlation of the two measures is just r = .54, p < .05. ?!?
Great video!
Why does the significance level has to be non-significant?
want to Know too ???
Thanks for the info =). However, one question: Is there any restriction with respect non-continous variables? Could this method be used to compare interval measures? Thanks!
Why would you not proceed with the Bland Altman Plot if the significance of the One-sample T-test was significant?
Please can I ask why you would not move forward with a Bland - Altman Plot if p value is statistically significant (i.e. 0.003)? What does a p value of 0.003 in this case mean? - Thank you, I know you made this video a while ago but it (any many others I've watched) have been very useful in my studies.
Did you find any answers to your question? I also obtained a p-value that was statistically significant and was wondering what it meant.
Thank u . But could u pls also elaborate the interpretation of the Bland Altman plot sir ?
THANK YOU
Nicely explain...I understood and implemented in my case....but could not able to conclude from results....Sir...can u help me in this regard
What if the dependant variables are categorical data (for example scores 1 to 9)? Can we still perform one sample t-test to see if there’s significant difference between test 1 and test 2?
Dear Dr. Tood, I was criticized by a refree for using one sample t-test before bland altman to make sure there was no statistically significant differences. So I have been looking for any literature to support my doing. I would be happy if you give me any advice.
Did you find any answers to your question? I was wondering the same thing.
Hi... can bland altman plot be applied for non normal distribution of data ?
Hello. Thank you for such an informative video. I don't if my plot is too maginified or what, only 4 dots can be seen on my graph despite there being several values. Regards.
why do you perform a one sample t test for this and not something else like a paired t?
In this case, a one-sample t-test of the differences yields the same results as a paired sample t-test.
Thank you very much for this video. However, I have my lower confidence interval falls below the limit of the plot and there is an SPSS message telling me that my lower confidence value is below the allowed limit. What could be the problem Dr. Grande please? I really started to get confused about how to solve this issue.
Looking forward to hearing a reply from you ASAP please.
Dr. Faraidoon Rawf
I also experienced the same problem. Did you solve this problem? i cannot draw the chart.
Might be slightly late... but you can double click to bring up the properties tab and go along to scale. From there you can enter the range of values for the Y axis.
Thanks
You're welcome!
Perfect
Thank you very much
You’re welcome!
hello dr todd
thank for this vedio . i have learned a lots ,,
i have a question sir ,, can we do bland-altman plot with 3 variables line .. and how we can do it ?
thanks a lots
i already did itraclass correlation reliability for theses 3 variables and now need to do, bland-altman plot
in my understanding of CI, it is based on SE and not on SD. how would you reconcile this Dr? Thank you. All the best. Thanks for uploading your videos
+Ed Marquez Confidence intervals are calculated using the SE for an unknown mean and an unknown standard deviation. The calculation in this video was based on an unknown mean and a known standard deviation, so the SD was used.
Actually... it's a sample SD and an estimate of sigma, the population standard deviation. Confidence intervals for sigma can be calculated from chi square distribution.
hiii sorry i was following ur video how did u came with 1.96
Same question, could you answer us please?
The screen numbers is not clearly visible
Thanks for the video. How do I get my chart to display the "0" line in the centre? the chart is set with the ranges of -5 to 15 and my (95%CI) ranges from LB: -8.08 to 11.00. So when I add the y-axis reference line for the lower limit...I get an error message saying it is beyond the limits of the chart!
Never mind, figured it out!
hey same happend with me . what did u do?
Harry Khan there was a setting to edit the upper and lower limits. Will have to get back on to SPSS and update later.
thanks perfect
You're welcome, thanks for watching -