This is one of the best You Tube tutorials on cross tabulation and Chi-square I had seen. Thank you for saving me. I have been confused for a week trying to understand how to report results.
I have done my masters in area related to this and this doesn’t make sense at all spending thousands of dollars to learn what you can learn with little bit of passion and an Internet connection.
Greetings from a new subscriber in Bali, Indonesia. Thank you very much for this clear and informative video from a fellow senior moment researcher. What I learned from the presentation is how to explain the findings of these tables, and why it is important to do so.
Very informative session confirming what I know and removing any doubts of what I had regarding the p value and its interpretation. Currently studying for a Masters in Public Health in this was quite helpful to watch. Thanks.
Thank you Sir for the clearest explanation of the p score I have ever heard. Your entire presentation was very clear. Do you talk further somewhere about how to determine significance if >20% of the cells are
Dear Sir, My hypothesis is " covid19 has no significant affect on income of the respondents.. The question from respondents is Have covid19 affected your income level? Yes No 250 answered yes while 50 not in one region while 250 yes and 30 No in other region... Used Chi square test, The p value is 0.02 less than 0.05. The question is " can I conclude that covid 19 significantly affected the income level of the repondents? Or There is significant difference b/w affected and non affected income of the respondents due to covid 19?
Dear Dr. Graham R Gibbs, Thank you for your clarity in explaining chi-square and it helps me to clear my confusions. I do have a doubt at 18.48 time, in the table rows of females have 3 cells with an expected value less than 5, and 1 cell although greater than 5 still low enough but your SPSS output at 19.18 time showed "0 cells have less than an expected value" and you got your result. Kindly clear my doubt, because I face this issue of low frequency count in some of my cells. Thanking you.
The cell number we need to worry about is the expected value. The table showing at 18.48 is the actual values - i.e. how many men or women were in each occupational group. The expected value is the number we would have got if thre were no difference between men and women and between the occupational groups. (There is an option in SPSS to display this in the table output, but I had opted not to show it in the table at 18.48.) As a general rule, where there are large numbers of individuals in a table (as there are in the Bank data) then the expected values of the cells for the small groupings (in this case the less common job categories) tend to be over 5 even if the actual values are very low. Indeed that is the case here. So we do not need to worry about the reliability of the chi square value because there are too many cells with expected values less than 5.
Hello Mr Graham, please I need you to help me with the formula on how to get the "Likelihood Ratio and the Linear by Linear Association" manually or through Excel because I don't have SPSS on my system and what am working on is urgent...please help out if it's possible
I really appreciate your prompt and extensive reply. Thanks to your videos that I understood the basics of chi-squared test. very helpful. yes I am going to report the findings as they are. but in the event that the sampling has a problem... does that discredit all my work. I had to fly to another country to get the data. my question is: does persistent failure to reject the null, means my data collection has a problem. in fact I also conducted interviews with actors and they reject the null can interviews be used as an evidence? Many Thanks Dr Graham.
+S. Khelifi It looks like you have lots of variables you are testing and none of them shows a significant result. In my experience that most often happens because the sample size is too small. There may be nothing else wrong with your data collection, just not enough of it. But you cannot exclude the possiblility that there really is no difference, so that is what you have to report. This becomes interesting if other research did find a difference.
Thank you Mr. Graham for this valuable video. I'm using crosstabs and chi-square to analyze the data of my survey but the assumption of no more than 20% of expected counts are less than 5 is violated for the majority of the variables. Is it statistically OK to weight cases for those variables in order to solve this problem?
Thank you so much for the video! could I take the p-value of fisher's exact test ( the gold standard) from the table instead of continuity correction? Many thanks!
Hello dear Dr. I have been given a questionnaire of 50 questions and my Dr has asked me to make a chi-square table for it. the questions are about teachers in class. could you give me a hint please , how to do it. I appreciate you in advance.
But I should also say that I'm trying to teach myself Statistics and SPSS and you are the first person who has helped me understand where the 'expected count' comes from. So thanks for that.
Mandy, you're right. Just before the end, at about 21.29 I do mention .001 in the displayed text and I talk about .005. I agree, this could be misleading and I think I was just getting tired by that point. I meant to say 0.05 and I should have pointed out that the actual probability calculated by SPSS was not only
Hallo, is it possible to also difference the cross tabs between woman and man? So if you also want to do the same analyse but see if it depends on the woman or man that answered or at the same time at the age? How do you do that? Thank you!!
Katrien Van den Bossche I'm not sure I really understand your question. We use the same table (and the same Chi square statistics) whether we are asking if women are more likely to be older than men or if the older students are more likely to be women than men.
Kleber Martins In this case I could have used Weight cases (using the frequencies that appear in each of the four cells of the contingency table as the weights). But these students would go on to work on a larger data set that contained many other variables and they would not have the contingency table to start with, but rather just the case by case values. So the approach I used here made more sense.
what if the p-value is always found to be more than alpha level. Does that mean that my sampling was not randomly selected? does this mean all my work is "to the flame"? I have almost finished my dissertation, so frustrating!
+S. Khelifi In that case you have no evidence for any relationship between the variables. I'm afraid that situation occurs all too often and it is frustrating. Why does it happen? In the first case it might be that there is, in reality, no relationship. So you are never going to find it. That's how the world is so you have to live with it. But you might have other evidence, perhaps from other research where such a relationship has been found. In that case there can be many reasons why your data don't show a similar significant relationship. One of the most common reasons is that your sample size is too small. But also, you might just be unlucky. Or you may have set up the data collection in a slightly different way so you don't get the same results or you may have made a mistake in data collection or data transcription etc. Of course, if other research suggests there should be a relationship, but after careful research you don't find such a relationship, that might be an interesting result. Maybe the original research was wrong, or maybe you have changed something significant in the way you did your research that changes things so that the relationship no longer holds. A null result is not a complete waste, but before you hold it up as a refutation of previous research you need to make sure that there is not some other explanation of the different results.
It is probably the result of rounding in the statistical program you are using. If the calculated p value is very small, e.g. 0.00001 the program may just display it as 0.000. So if you have this result then the result is significant at alphas of both 0.05 and 0.01. Beware, though, this often happens when you have a very large sample. It means you could not have got that result by mere chance, there is something going on, but it may not be analytically important. That depends on the hypothesis you are testing.
You can use chi square on a 2x4 table - or any size table for that matter. You don't then need the continuity correction that is just for 2x2 tables. However, with larger tables there is more chance of having cells with expected values of
This is one of the best You Tube tutorials on cross tabulation and Chi-square I had seen. Thank you for saving me. I have been confused for a week trying to understand how to report results.
Still the best tutorial at 2024. It help me a lot
You just explained to me (for FREE) what my $6,000 a quarter master's degree program SHOULD have explained! THANK YOU!
Doing my Masters as well and I agree with you! lol
I have done my masters in area related to this and this doesn’t make sense at all spending thousands of dollars to learn what you can learn with little bit of passion and an Internet connection.
We learn this at undergraduate.
Greetings from a new subscriber in Bali, Indonesia. Thank you very much for this clear and informative video from a fellow senior moment researcher. What I learned from the presentation is how to explain the findings of these tables, and why it is important to do so.
you seem like a really sweet man, very comprehensive lecture. thank you ! i cried a little, exams are making me breakdown a little
I paid 9,500 k for this year and learned nothing, this essentially free video thought me more than my uni. Thanks!
Your lecture is superb! Explain everything clearly; very helpful during my dissertation time. Thank you so much!
Your videos are extremely helpful! Brilliant lectures, thank you very much for putting them up here for everyone to learn from!
The most valuable info for me was the interpretation and I thank you for this sir.
Very informative session confirming what I know and removing any doubts of what I had regarding the p value and its interpretation. Currently studying for a Masters in Public Health in this was quite helpful to watch. Thanks.
Exceptional Work!
You basically taught me the whole semester, thank you :D
If that is a whole semester in 20 mins ..then I'd hate to be at the college u are at! I hope they give you your tuition fees back! 😂
Thank you for the clear lecture, I will certainly share. Best wishes!
Thank you Sir for the clearest explanation of the p score I have ever heard. Your entire presentation was very clear. Do you talk further somewhere about how to determine significance if >20% of the cells are
to the point after hours of surfing !!! thanks a lot!!
Sir, your video is really helpful. A true legend TY
Thank you for this great lecture! It was very helpful indeed!
It was of great help Sir! Thankyou so much for making it.
Thank you, Sir. Very well explained.
Thanks for the lecture.
Thank you very much. Your class enlightened me.
Thank you!! Your video helped me so much.
Thank you for clear and informative lecture
Enjoyed it
Thats the best explanation ever! thank u!
Great thanks
Great
This was excellent. Thank you.
Dear Sir,
My hypothesis is
" covid19 has no significant affect on income of the respondents..
The question from respondents is
Have covid19 affected your income level?
Yes
No
250 answered yes while 50 not in one region while 250 yes and 30 No in other region...
Used Chi square test,
The p value is 0.02 less than 0.05.
The question is
" can I conclude that covid 19 significantly affected the income level of the repondents?
Or
There is significant difference b/w affected and non affected income of the respondents due to covid 19?
Thank you!
Dear Dr. Graham R Gibbs, Thank you for your clarity in explaining chi-square and it helps me to clear my confusions. I do have a doubt at 18.48 time, in the table rows of females have 3 cells with an expected value less than 5, and 1 cell although greater than 5 still low enough but your SPSS output at 19.18 time showed "0 cells have less than an expected value" and you got your result. Kindly clear my doubt, because I face this issue of low frequency count in some of my cells. Thanking you.
The cell number we need to worry about is the expected value. The table showing at 18.48 is the actual values - i.e. how many men or women were in each occupational group. The expected value is the number we would have got if thre were no difference between men and women and between the occupational groups. (There is an option in SPSS to display this in the table output, but I had opted not to show it in the table at 18.48.) As a general rule, where there are large numbers of individuals in a table (as there are in the Bank data) then the expected values of the cells for the small groupings (in this case the less common job categories) tend to be over 5 even if the actual values are very low. Indeed that is the case here. So we do not need to worry about the reliability of the chi square value because there are too many cells with expected values less than 5.
@@GrahamRGibbs Thank you sir
Thank You...great teaching.
Thank you no Thai professor makes me understand it as well as what you got here .
Hello Mr Graham, please I need you to help me with the formula on how to get the "Likelihood Ratio and the Linear by Linear Association" manually or through Excel because I don't have SPSS on my system and what am working on is urgent...please help out if it's possible
Thank you for Sharing, Which is helpful to me.
Thank you so much Sir it really help me in my analysis
Thanks a lot.However there is need to display and explain key steps of keying in data when using a chi-square.
very good work
GREAT!!!
You ROCK!!
I really appreciate your prompt and extensive reply. Thanks to your videos that I understood the basics of chi-squared test. very helpful. yes I am going to report the findings as they are. but in the event that the sampling has a problem... does that discredit all my work. I had to fly to another country to get the data. my question is: does persistent failure to reject the null, means my data collection has a problem. in fact I also conducted interviews with actors and they reject the null can interviews be used as an evidence? Many Thanks Dr Graham.
+S. Khelifi It looks like you have lots of variables you are testing and none of them shows a significant result. In my experience that most often happens because the sample size is too small. There may be nothing else wrong with your data collection, just not enough of it. But you cannot exclude the possiblility that there really is no difference, so that is what you have to report. This becomes interesting if other research did find a difference.
Thank you Mr. Graham for this valuable video. I'm using crosstabs and chi-square to analyze the data of my survey but the assumption of no more than 20% of expected counts are less than 5 is violated for the majority of the variables. Is it statistically OK to weight cases for those variables in order to solve this problem?
Does anyone know how to interpret a result if you were to perform a (Chi) χ2 test and obtain a p-value of 0.005.
what happens if the expected count in the 2 by 2 table is more than 20%
Thanks for the explanation .Any one knows what is the music in the beginning of the video :)?
Thank you so much for the video! could I take the p-value of fisher's exact test ( the gold standard) from the table instead of continuity correction? Many thanks!
Hello dear Dr.
I have been given a questionnaire of 50 questions and my Dr has asked me to make a chi-square table for it. the questions are about teachers in class. could you give me a hint please , how to do it. I appreciate you in advance.
Right at the very end of the clip you state that p< 0.001 which looks to be true but I think this should say p
But I should also say that I'm trying to teach myself Statistics and SPSS and you are the first person who has helped me understand where the 'expected count' comes from. So thanks for that.
Mandy, you're right. Just before the end, at about 21.29 I do mention .001 in the displayed text and I talk about .005. I agree, this could be misleading and I think I was just getting tired by that point. I meant to say 0.05 and I should have pointed out that the actual probability calculated by SPSS was not only
Mandy Aspland
Mandy, I've now annotated the video to make this clear for the eagle eyed, like you. Thanks for pointing this out.
Hallo, is it possible to also difference the cross tabs between woman and man?
So if you also want to do the same analyse but see if it depends on the woman or man that answered or at the same time at the age?
How do you do that?
Thank you!!
Katrien Van den Bossche I'm not sure I really understand your question. We use the same table (and the same Chi square statistics) whether we are asking if women are more likely to be older than men or if the older students are more likely to be women than men.
Hello dear Graham,
Why did not you use Weight case ?
Best regards,
Dr. Kleber
Kleber Martins In this case I could have used Weight cases (using the frequencies that appear in each of the four cells of the contingency table as the weights). But these students would go on to work on a larger data set that contained many other variables and they would not have the contingency table to start with, but rather just the case by case values. So the approach I used here made more sense.
what if the p-value is always found to be more than alpha level. Does that mean that my sampling was not randomly selected? does this mean all my work is "to the flame"? I have almost finished my dissertation, so frustrating!
+S. Khelifi In that case you have no evidence for any relationship between the variables. I'm afraid that situation occurs all too often and it is frustrating. Why does it happen? In the first case it might be that there is, in reality, no relationship. So you are never going to find it. That's how the world is so you have to live with it.
But you might have other evidence, perhaps from other research where such a relationship has been found. In that case there can be many reasons why your data don't show a similar significant relationship. One of the most common reasons is that your sample size is too small. But also, you might just be unlucky. Or you may have set up the data collection in a slightly different way so you don't get the same results or you may have made a mistake in data collection or data transcription etc.
Of course, if other research suggests there should be a relationship, but after careful research you don't find such a relationship, that might be an interesting result. Maybe the original research was wrong, or maybe you have changed something significant in the way you did your research that changes things so that the relationship no longer holds.
A null result is not a complete waste, but before you hold it up as a refutation of previous research you need to make sure that there is not some other explanation of the different results.
great great video
1400
Sir, my all likelihood p-values came to be p=0, what does it signifies? is it significant at 0.05 and 0.01 too?
It is probably the result of rounding in the statistical program you are using. If the calculated p value is very small, e.g. 0.00001 the program may just display it as 0.000. So if you have this result then the result is significant at alphas of both 0.05 and 0.01. Beware, though, this often happens when you have a very large sample. It means you could not have got that result by mere chance, there is something going on, but it may not be analytically important. That depends on the hypothesis you are testing.
@@GrahamRGibbs Thankyou so much sir for explaining it so well!
Which test should I use for a 2x4 table ?
You can use chi square on a 2x4 table - or any size table for that matter. You don't then need the continuity correction that is just for 2x2 tables. However, with larger tables there is more chance of having cells with expected values of
Thank you!