Could you comment on the p value for the correlation? and what it means then if you do find a statistically significant correlation but a non significant paired t test
I have a doubt ..When you conclude in apa style that clicker training is more effective than traditional training as we were doing two tailed test where we can only conclude that they are different by rejecting null hypothesis if we were using one tailed test then we can mention like one is more effective than other what's the difference I couldn't understand
As I understand, paired samples T-test is used traditionally for before and after design. Here we have two separate groups of dogs assigned to either clicker or food training program. They, as I understand, are not related except for the fact that they are simply dogs :) So my question would be why we are using paired samples t test here? Why are they related if none of the dogs had both training programs passed? @Research By Design
This same paired t test can be used if (a) sores from the same people are paired in time, such as before/after, or (b) scores from related subjects are paired. In this example, because the puppies are "twins" in essence, their scores cannot be assumed to be independent. We pair the twins to account for the non-independence. Hope that clarifies. Thanks for the question!
Hello!I have a question for you. You used the same data in two methods, paired sample t-test and on the previous video independet sample t-test and you concluded on different things. With the paired samle t-test you say that the difference in the means is significant but with the independent sample t-test there was no significant difference. How is that possible?Am I missing something?Thank you for your time!
Yes, that is exactly right! The point is that repeated measures (i.e. within subjects design, paired samples) have more "power" to detect differences than independent measures. If you look at the video on independent samples t test, I also discuss this. I designed this example so that the same data would give a non-significant independent t test, but a significant paired t test to make this point. So, no, you did not miss anything...in fact you got the main idea loud and clear. Thanks for the comment!
Sir @ Research By Design, I am new to this stats thing, Could you please answer my questions? First, What if the t value is negative? how we are going to compare it with the critical value? and my second question is you have said "We also see the 95% confidence interval for the difference between paired sample means shows a lower bound of +1.78 and an upper bound of 2.322. Because this confidence interval does not cross 0". But sir here both +1.78 and 2.32 greater than 0 that means they have crossed zero. how you have interpreted that they have not crossed zero?
No problem...(1) negative t value is interpreted the same as positive, the sign only indicates the direction of change, (2) the lower and upper CI "cross zero" when one is negative and the other is positive (therefore their range includes 0). In this case, they are both on the same side of zero. Hope that helps.
If the p = .053, then the test would be non-significant. That said, we should also consider the sample size and effect size. Honestly, adding one more participant might be enough to flip over to significance. And adding five more might be non-significant again, and 10 more would be significant...so we interpret the effect size to get a more stable estimate. The nature of p values is that they flip in and out of significance before they finally converge. May be more than you were asking about, but there you go. Good luck!
Yes, there is a way...probably too much to explain in a comment. Try doing the paired t test as you see here, but then look up the one-sample t critical value in a t-table from a statistics text book (or google one online). That is the quickest method. Good luck.
You could use this same test with a different pair of variables, but if you have three (or more) variables that you want to compare (say, a pretest, posttest, and 3-mo follow-up), then you would use a repeated measures ANOVA.
Could you comment on the p value for the correlation? and what it means then if you do find a statistically significant correlation but a non significant paired t test
Those are two different tests (correlation vs. t-test). A significant correlation MAY even be a problem if the variables are so intercorrelated that they are measuring the same thing. Both may be significant, or only one, or neither. You do not change the way that you interpret the t-test based upon the correlation. Also, the Cohen's d will be affected by intercorrelation.
Best SPSS tutorial video I have ever seen. Thank you so much!
Lucid explanation...well planned...well organised...Great work!!!
Glad you liked it!
very helpful video! good explanation, everything is clear. thank you!
Thanks for the comment and that you found the video helpful.
Perfect!
Very Good!
Could you comment on the p value for the correlation? and what it means then if you do find a statistically significant correlation but a non significant paired t test
I have a doubt ..When you conclude in apa style that clicker training is more effective than traditional training as we were doing two tailed test where we can only conclude that they are different by rejecting null hypothesis if we were using one tailed test then we can mention like one is more effective than other what's the difference I couldn't understand
I got it 👍
thank you so much
As I understand, paired samples T-test is used traditionally for before and after design. Here we have two separate groups of dogs assigned to either clicker or food training program. They, as I understand, are not related except for the fact that they are simply dogs :) So my question would be why we are using paired samples t test here? Why are they related if none of the dogs had both training programs passed? @Research By Design
This same paired t test can be used if (a) sores from the same people are paired in time, such as before/after, or (b) scores from related subjects are paired. In this example, because the puppies are "twins" in essence, their scores cannot be assumed to be independent. We pair the twins to account for the non-independence. Hope that clarifies. Thanks for the question!
What is cohens d... Never heard of it.. Why is it important?
Hello!I have a question for you. You used the same data in two methods, paired sample t-test and on the previous video independet sample t-test and you concluded on different things. With the paired samle t-test you say that the difference in the means is significant but with the independent sample t-test there was no significant difference. How is that possible?Am I missing something?Thank you for your time!
Yes, that is exactly right! The point is that repeated measures (i.e. within subjects design, paired samples) have more "power" to detect differences than independent measures. If you look at the video on independent samples t test, I also discuss this. I designed this example so that the same data would give a non-significant independent t test, but a significant paired t test to make this point. So, no, you did not miss anything...in fact you got the main idea loud and clear. Thanks for the comment!
Sir @
Research By Design, I am new to this stats thing, Could you please answer my questions?
First, What if the t value is negative? how we are going to compare it with the critical value? and my second question is you have said "We also see the 95% confidence interval for the difference between paired sample means shows a lower bound of +1.78 and an upper bound of 2.322. Because this confidence interval does not cross 0". But sir here both +1.78 and 2.32 greater than 0 that means they have crossed zero. how you have interpreted that they have not crossed zero?
No problem...(1) negative t value is interpreted the same as positive, the sign only indicates the direction of change, (2) the lower and upper CI "cross zero" when one is negative and the other is positive (therefore their range includes 0). In this case, they are both on the same side of zero. Hope that helps.
@@ResearchByDesign Ok Okay.... Now I got your point. Thanks for your time.
Sir if my two tailed value is 0.053 so is it significant???? What so we consider?
If the p = .053, then the test would be non-significant. That said, we should also consider the sample size and effect size. Honestly, adding one more participant might be enough to flip over to significance. And adding five more might be non-significant again, and 10 more would be significant...so we interpret the effect size to get a more stable estimate. The nature of p values is that they flip in and out of significance before they finally converge. May be more than you were asking about, but there you go. Good luck!
@@ResearchByDesign thnxs alot....
What if I want a one-tailed paired sample test?? Is there no way??
Yes, there is a way...probably too much to explain in a comment. Try doing the paired t test as you see here, but then look up the one-sample t critical value in a t-table from a statistics text book (or google one online). That is the quickest method. Good luck.
@@ResearchByDesign Thanks!
How did you get the df as 7?
The df for a paired samples t-test is the number of pairs minus 1. I used 8 pairs of dogs, so df = 7. Cheers
do this test can be use if we have more that one pair?
You could use this same test with a different pair of variables, but if you have three (or more) variables that you want to compare (say, a pretest, posttest, and 3-mo follow-up), then you would use a repeated measures ANOVA.
@@ResearchByDesign great, I will try first using ANOVA. thank you so much sir
Could you comment on the p value for the correlation? and what it means then if you do find a statistically significant correlation but a non significant paired t test
Those are two different tests (correlation vs. t-test). A significant correlation MAY even be a problem if the variables are so intercorrelated that they are measuring the same thing. Both may be significant, or only one, or neither. You do not change the way that you interpret the t-test based upon the correlation. Also, the Cohen's d will be affected by intercorrelation.