You have to look it up in a table. Most statistical textbooks have it in an appendix. I recopied some of the one I use, but I can't show it for copyright reasons.
Doesn't q have something to do with the range? Like the range of the set of sample means (not individual pairs of sample means)? I am trying to learn this topic and have seen that indicated in other places.
@@marka2773 Most statistics program apply a correction/approximation technique. R, the stats program I use, uses the Kramer method I believe. To be honest, I've never had to hand calculate it, and don't know as much as I should on this specific topic. Given that Tukeys is a modified t-test, I'd bet the correction looks similar to how an independent samples t-test uses pooled variance. You put MS-error over the n for each group and add them together. I did a little digging and this page by David Howell at UVM says that something along those lines is the case: www.uvm.edu/~statdhtx/StatPages/MultipleComparisons/unequal_ns_and_mult_comp.html Here is an older paper about the subject: link.springer.com/article/10.1007/BF02291420
Its still Tukey's HSD. It is just a different way of going about it. Tukey's HSD is really just a t-test where you correct for multiple comparisons and use the MS within from the ANOVA to get a common standard error.
To solve your specific problem, I would use a statistics program to look up a value from a given distribution. You might be able to do it in excel. I know it can do this for the t, f, and I'm pretty sure Chi-squared as well in excel. My go to program anymore is R. If you can use it here is a link to a stackexchange article about generating custom q tables (stats.stackexchange.com/questions/177386/how-to-obtain-tukey-table-in-r). When I ran the code qtukey(p = 0.95, nmeans = 226, df = 54) it gave be a value of 7.071167.
Tough to say without more information. Bonferroni corrected t-tests are always a safe, conservative, and defensible option. Tukey's HSD is safe if you have a between subjects experiment and it is not as conservative as Bonferroni corrected tests. However, if it is repeated measures and assumption of sphericity is not met, it can be problematic. That is why SPSS won't let you use Tukey's as a post hoc test. Do NOT use LSD or Newman-Keuls. LSD is not acceptable when there are over three comparisons, and Newman-Keuls has been shown to inadequately control for type I error.
Based on my poor translation, I think you figured it out. I'm sorry I can't show the table to demo it. I'm not sure whether or not it would violate copyright laws.
This is the second of 2 videos (1st one calculates the ANOVA). The dependent variable of driving simulator performance and the manipulation of the independent variable (phone use while driving) was stated at the start of the last video: ua-cam.com/video/OikHQKviGpo/v-deo.html
How did you get the 3.77 critical point on the tukey q table? I'm a bit confused on how to interpret the table 10:52
You have to look it up in a table. Most statistical textbooks have it in an appendix. I recopied some of the one I use, but I can't show it for copyright reasons.
Doesn't q have something to do with the range? Like the range of the set of sample means (not individual pairs of sample means)? I am trying to learn this topic and have seen that indicated in other places.
how did you obtain n=5? are you just picking any n from each of the groups or are you averaging them?
All groups have the same n of 5.
@@s.wesleybeckwith3561 What if my groups have different n values?
@@marka2773 Most statistics program apply a correction/approximation technique. R, the stats program I use, uses the Kramer method I believe. To be honest, I've never had to hand calculate it, and don't know as much as I should on this specific topic.
Given that Tukeys is a modified t-test, I'd bet the correction looks similar to how an independent samples t-test uses pooled variance. You put MS-error over the n for each group and add them together. I did a little digging and this page by David Howell at UVM says that something along those lines is the case: www.uvm.edu/~statdhtx/StatPages/MultipleComparisons/unequal_ns_and_mult_comp.html
Here is an older paper about the subject: link.springer.com/article/10.1007/BF02291420
@@s.wesleybeckwith3561 Thank you both. I was searching the answer to this topic.
What is the name of the test you are doing initially? I mean where you are calculating q using the formula similar to T - test?
Its still Tukey's HSD. It is just a different way of going about it. Tukey's HSD is really just a t-test where you correct for multiple comparisons and use the MS within from the ANOVA to get a common standard error.
Can you please suggest how to get q table values when df=54 and k=226? I found that the q table is limited to k=20. Please help me, thank you.
To solve your specific problem, I would use a statistics program to look up a value from a given distribution. You might be able to do it in excel. I know it can do this for the t, f, and I'm pretty sure Chi-squared as well in excel.
My go to program anymore is R. If you can use it here is a link to a stackexchange article about generating custom q tables (stats.stackexchange.com/questions/177386/how-to-obtain-tukey-table-in-r). When I ran the code qtukey(p = 0.95, nmeans = 226, df = 54) it gave be a value of 7.071167.
And in excel...
www.real-statistics.com/students-t-distribution/studentized-range-distribution/
@@s.wesleybeckwith3561
Thank you for kind help, R Codes that you gave me through the above link are giving results.
What post hoc test should i use if i have 4 groups/treatments??please answer! We have a defense next week and im not yet done with this thank you!!
Tough to say without more information. Bonferroni corrected t-tests are always a safe, conservative, and defensible option. Tukey's HSD is safe if you have a between subjects experiment and it is not as conservative as Bonferroni corrected tests. However, if it is repeated measures and assumption of sphericity is not met, it can be problematic. That is why SPSS won't let you use Tukey's as a post hoc test.
Do NOT use LSD or Newman-Keuls. LSD is not acceptable when there are over three comparisons, and Newman-Keuls has been shown to inadequately control for type I error.
How did you get the n=5
"n" is equal to the number of scores in each condition. There were 5 people in the no phone condition (and the others) and so n=5.
How do you obtain the 3.77, the critical point?
Ya vi.... Tabla de puntos porcentuales rango estunderizado (5%)
q(0.05; K, N-K) = 3.77 | Tabla A5
Based on my poor translation, I think you figured it out. I'm sorry I can't show the table to demo it. I'm not sure whether or not it would violate copyright laws.
too bad you didn't state what each variable was instead of just reading out the letters.
This is the second of 2 videos (1st one calculates the ANOVA). The dependent variable of driving simulator performance and the manipulation of the independent variable (phone use while driving) was stated at the start of the last video: ua-cam.com/video/OikHQKviGpo/v-deo.html
@@s.wesleybeckwith3561 Thanks!