Hi David, Thank you so much for your videos they have been so helpful during my master in communication science. I am now analyzing survey data for my master thesis and have a few questions. I am in way over my head I would appreciate your help so much. Here is the situation:
I have three variables, covid stress (used Covid stress scales measure), mood regulation strategies(there are 7)(used Brief Mood in Music Regulation measure) and music genre(14 genres)(used STOMP short test of music preferences measure). I two hypotheses: H1: Individuals with higher levels of perceived stress due to the pandemic will report more use of BMMR strategies that focus on dealing with negative feelings (solace, Diversion, and Discharge) H2: BMMR strategies that focus on dealing with negative feelings solace, diversion, and discharge will be associated with more frequent listening of classical (for solace), soul/funk (for diversion), and heavy metal (for discharge)
- I have totaled the covid stress scales score and created a new variable with the sum of all for each respondent, I believe this variable is not good to go for analysis - Emotion regulation strategies - 21 liker scale items, 3 items per strategy, there are seven strategies, I have taken the mean of each set of three and created a new variable for each strategy (there are seven) - likert scale preferences for 14 genres of music
Question: I believe I need to do a regression for H1 with covid stress as IV and mood regulation strategies as DV and for H2 a regression with the three mood regulation strategies of , , and as IV and the 14 music genres as DV, is this correct?
Question: Do I do a separate regression for each sub scale of the BMMR? So there would be seven regressions? For H2 do I need to do 14 regressions for the 14 music genres?
@@PhiloSofieMusic For H1, you could do three simple regression analyses, one per strategy (DV) and with stress as the IV in each. You could also do three simple regression analyses for H2, with a different type of strategy as the IV in each and a different genre as the DV in each. I'm saying three as in your hypotheses above you seem to refer to three relationships between strategy and stress and three relationships between strategy and genre, though you could do as many regressions as you need if you want to examine more relationships. "Simple" regression analyses will always have one IV and one DV, whereas "multiple" regression analyses can have multiple IVs though will still have one DV.
Hello quick question regarding a different subject. What would be required if one levenes test comes back not significant and the other comes back significant?
Hi Harry, thanks for your question. It depends on what tests you're planning to run. E.g., if you're running t-tests, you would just need to refer to different parts of the t-test output tables for each test, whereas for ANOVAs you might consider using a more stringent alpha (e.g., .01 instead of .05) for the ANOVA associated with the significant Levene's test.
Hello Dr. David! Nice video! I wish to ask you a question regarding skewness and kurtosis and I would appreciate it if you can help! I have scores of females and males and I want to compare them using independent t test so I should verify normality condition. In this case, if I do skewness and kurtosis for each gender score is it considered the same as doing the skewness and kurtosis of all the scores? I mean if the values of the skewness and kurtosis for each gender score were in the range of normality does that imply that the data is normal or we should verify the skewness and kurtosis of the whole scores?
Hi Katy, it’s usually just assumed that the data are normally distributed within each group (i.e., it’s not necessary to check that scores within the whole dataset are normally distributed).
Hi David, Hope you are doing good :) Can I add a suggestion to your list of future videos? I would really appreciate a video on Cluster Analysis Thanks!
Hello! Your video was very helpful. I have one question though. If I got it right, when I have 2 indipendent groups and I want to compare them in order to assess if they are significantly different from each other, I will first run a normality test in order to decide if I will proceed with a T-test or a Mann-Whitney U test. My first group consists of 64 samples and my second group consists of 16 samples. The normality test should be run in all my data (meaning I will consider my two groups as one and I will run a normality test to all of them) or should I run a normality test to each of my groups? Also, If I run a normality test to each of my groups and my first group (n=64) has normal distribution and the second one (n=16) does not have normal distribution, should I proceed with a T-test or with a Mann-Whitney U? Maybe should I choose Mann-Whitney U because my first group is not normal distributed and my second group is small?
Hi, thanks for your question and sorry for the delayed response. Yes, if either of the groups has a non-normal distribution, you would normally use a Mann-Whitney test instead of a t-test.
basic question - i tried this but in my output, where you are having the Sig column, followed by a numerical value only, I am having the Sig column but
@@DavidRobinsonPhD So it means my DV is statistically significant and therefore "unnormal" - exactly what are do not want in this case right? However my point is, if we ran the same data, you would have the Sig column and a numerical value only under it (say .067) whereas I would have a Sig column and a
@@wtc4752 Yes, < .001 suggests the data are significantly different from normal, which is usually not what you want if you're planning to run parametric tests. I'm not aware of any way of changing the settings in the way you're referring to. Have you tried running the test with data that you know are normally distributed (in which case you would expect to see an exact numerical value rather than < .001)?
For assessing normality on the paired samples t-test (with pre test and post-test scores), should the descriptions > explore be run on the pre test and post-test individual variables or a difference score (pre - post)?
Hi, thanks for your question. The paired t-test assumes that the differences between pre and post scores are normally distributed, so a variable representing differences should be used.
Hi David, Thank you so much for your videos they have been so helpful during my master in communication science. I am now analyzing survey data for my master thesis and have a few questions. I am in way over my head I would appreciate your help so much. Here is the situation:
I have three variables, covid stress (used Covid stress scales measure), mood regulation strategies(there are 7)(used Brief Mood in Music Regulation measure) and music genre(14 genres)(used STOMP short test of music preferences measure). I two hypotheses: H1: Individuals with higher levels of perceived stress due to the pandemic will report more use of BMMR strategies that focus on dealing with negative feelings (solace, Diversion, and Discharge)
H2: BMMR strategies that focus on dealing with negative feelings solace, diversion, and discharge will be associated with more frequent listening of classical (for solace), soul/funk (for diversion), and heavy metal (for discharge)
- I have totaled the covid stress scales score and created a new variable with the sum of all for each respondent, I believe this variable is not good to go for analysis
- Emotion regulation strategies - 21 liker scale items, 3 items per strategy, there are seven strategies, I have taken the mean of each set of three and created a new variable for each strategy (there are seven)
- likert scale preferences for 14 genres of music
Question: I believe I need to do a regression for H1 with covid stress as IV and mood regulation strategies as DV and for H2 a regression with the three mood regulation strategies of , , and as IV and the 14 music genres as DV, is this correct?
Question: Do I do a separate regression for each sub scale of the BMMR? So there would be seven regressions? For H2 do I need to do 14 regressions for the 14 music genres?
@@PhiloSofieMusic For H1, you could do three simple regression analyses, one per strategy (DV) and with stress as the IV in each. You could also do three simple regression analyses for H2, with a different type of strategy as the IV in each and a different genre as the DV in each. I'm saying three as in your hypotheses above you seem to refer to three relationships between strategy and stress and three relationships between strategy and genre, though you could do as many regressions as you need if you want to examine more relationships. "Simple" regression analyses will always have one IV and one DV, whereas "multiple" regression analyses can have multiple IVs though will still have one DV.
Hello quick question regarding a different subject. What would be required if one levenes test comes back not significant and the other comes back significant?
Hi Harry, thanks for your question. It depends on what tests you're planning to run. E.g., if you're running t-tests, you would just need to refer to different parts of the t-test output tables for each test, whereas for ANOVAs you might consider using a more stringent alpha (e.g., .01 instead of .05) for the ANOVA associated with the significant Levene's test.
Hello Dr. David! Nice video!
I wish to ask you a question regarding skewness and kurtosis and I would appreciate it if you can help!
I have scores of females and males and I want to compare them using independent t test so I should verify normality condition. In this case, if I do skewness and kurtosis for each gender score is it considered the same as doing the skewness and kurtosis of all the scores?
I mean if the values of the skewness and kurtosis for each gender score were in the range of normality does that imply that the data is normal or we should verify the skewness and kurtosis of the whole scores?
Hi Katy, it’s usually just assumed that the data are normally distributed within each group (i.e., it’s not necessary to check that scores within the whole dataset are normally distributed).
@@DavidRobinsonPhD Thank you very much!
Would you do the same thing, as in 3:27 - 5:55, if you had 5 different groups?
Hi David,
Hope you are doing good :)
Can I add a suggestion to your list of future videos?
I would really appreciate a video on Cluster Analysis
Thanks!
Hi, thanks for the suggestion, I'll add that to the list for future videos!
Hello! Your video was very helpful. I have one question though. If I got it right, when I have 2 indipendent groups and I want to compare them in order to assess if they are significantly different from each other, I will first run a normality test in order to decide if I will proceed with a T-test or a Mann-Whitney U test. My first group consists of 64 samples and my second group consists of 16 samples. The normality test should be run in all my data (meaning I will consider my two groups as one and I will run a normality test to all of them) or should I run a normality test to each of my groups? Also, If I run a normality test to each of my groups and my first group (n=64) has normal distribution and the second one (n=16) does not have normal distribution, should I proceed with a T-test or with a Mann-Whitney U? Maybe should I choose Mann-Whitney U because my first group is not normal distributed and my second group is small?
Hi, thanks for your question and sorry for the delayed response. Yes, if either of the groups has a non-normal distribution, you would normally use a Mann-Whitney test instead of a t-test.
basic question - i tried this but in my output, where you are having the Sig column, followed by a numerical value only, I am having the Sig column but
If you see
@@DavidRobinsonPhD
So it means my DV is statistically significant and therefore "unnormal" - exactly what are do not want in this case right?
However my point is, if we ran the same data, you would have the Sig column and a numerical value only under it (say .067) whereas I would have a Sig column and a
@@wtc4752 Yes, < .001 suggests the data are significantly different from normal, which is usually not what you want if you're planning to run parametric tests.
I'm not aware of any way of changing the settings in the way you're referring to. Have you tried running the test with data that you know are normally distributed (in which case you would expect to see an exact numerical value rather than < .001)?
nice, please could you add an index(MPI, HDI, HCI) calculation videos using Excel or SPSS, Please?
Hi, thanks for the suggestion. I'll add that to my list of future videos.
@@DavidRobinsonPhD please, as early as possible.
For assessing normality on the paired samples t-test (with pre test and post-test scores), should the descriptions > explore be run on the pre test and post-test individual variables or a difference score (pre - post)?
Hi, thanks for your question. The paired t-test assumes that the differences between pre and post scores are normally distributed, so a variable representing differences should be used.
Hi David, do you have an email where I can contact you about my results?