Thanks for the information. Is it acceptable to use recoded Likert scale data for the chi-square test? Is it appropriate to use the Kruskal Wallis test for assessing association? My understanding is that it's typically used for detecting differences.
Mine keeps saying no statistics can be run because one of my variables is a constant. But I used the same variable to find significant between it and others.
Dear Anthony, thanks for the great video. If we have a 2x4 situation, can we still use crosstabs to test for a "trend"? Situation: Outcome: Death/Alive (binary) Independent variables: Patient took Medicine A/B/C/D
Dear Takanna, thank you for replying! Yes, you can! It will look a little complicated but it can be run, just be aware that with multiple variables to ensure you are meeting the requirements of expected count greater than 5. As you're adding in extra sometimes your count will be lower as you're splitting across multiples. But yes in short you can, good luck!
Then unfortunately the test is not valid and you cannot run any inferential statistics, you won't have enough data to make any valid statistically significant findings. Instead you would need to rely on percentages which is okay, you just can't use words like statistically significant, or Group A is more likely than Group B. However if you have male and female and yes and no so a 2 x 2 table, then you can use a fisher test. If your data is eligible for that, SPSS will display a fisher test in the box where the chi-square test is. Same principle, less than 0.05 is significant. However caution as a Fisher test is not as accurate and has more margin or error.
Thanks William! Sure, for that you need to run a Mann-Whitney U or a Kruskall-Wallis H test depending on how many variables you have. I'll link those videos down below for you.
If I have the same variable in the same population (qualitative, like smoking habit: 1:yes 2:no 3: ex) and I want to see how it evolves on time (baseline-12 months-24 months) can I use Chi-Square?
Thanks for the information.
Is it acceptable to use recoded Likert scale data for the chi-square test?
Is it appropriate to use the Kruskal Wallis test for assessing association? My understanding is that it's typically used for detecting differences.
Mine keeps saying no statistics can be run because one of my variables is a constant. But I used the same variable to find significant between it and others.
Dear Anthony, thanks for the great video. If we have a 2x4 situation, can we still use crosstabs to test for a "trend"?
Situation:
Outcome: Death/Alive (binary)
Independent variables: Patient took Medicine A/B/C/D
Dear Takanna, thank you for replying! Yes, you can! It will look a little complicated but it can be run, just be aware that with multiple variables to ensure you are meeting the requirements of expected count greater than 5. As you're adding in extra sometimes your count will be lower as you're splitting across multiples. But yes in short you can, good luck!
@@AnthonyCliffePHD thank u for the reply! im blown away!
would u consider making a video to explain the above too?
If I have time I will :)
This helped me a lot thankyou :)
but what happens if you violate the assumptions and cannot recode your data? i.e. i just have fewer pp in a comparison group
Then unfortunately the test is not valid and you cannot run any inferential statistics, you won't have enough data to make any valid statistically significant findings. Instead you would need to rely on percentages which is okay, you just can't use words like statistically significant, or Group A is more likely than Group B. However if you have male and female and yes and no so a 2 x 2 table, then you can use a fisher test. If your data is eligible for that, SPSS will display a fisher test in the box where the chi-square test is. Same principle, less than 0.05 is significant. However caution as a Fisher test is not as accurate and has more margin or error.
@@AnthonyCliffePHD Thank you very much! This is extremely helpful for my thesis :)
Thank you for this video.
I'd like to know how to test association between variables measured on a likert scale
Thanks William! Sure, for that you need to run a Mann-Whitney U or a Kruskall-Wallis H test depending on how many variables you have. I'll link those videos down below for you.
Kruskall test: ua-cam.com/video/8JCHhGRWXuM/v-deo.html
Mann Whitney: ua-cam.com/video/C8sIn5L-Y4o/v-deo.html
If I have the same variable in the same population (qualitative, like smoking habit: 1:yes 2:no 3: ex) and I want to see how it evolves on time (baseline-12 months-24 months) can I use Chi-Square?
Sorry only just seen this so probably too late for you but you're best using a friedman rank test to compare groups over time
Haii, is it the same method for recode on column if i have more than 3 cells? thank you
Yes, you can have as many cells in your column and recode them down to 2
So males and females do not differ in the whether they are informed or not?
Yep!
Thank you so muchhhhhhg
My pleasure thank you for watching!
L😊😮p