I am very grateful you make these videos, you are a real savior. Could you please give a reference for rule 1 and rule 2 of normality please, I need to mention that in my writing? And more importantly, How we interpret the result of the transformed data??? it looks complicated, especially when we use log transformation. Many Thanks.
Thanks a lot for your prompt reply. Please can you help me on the following questions as well? 1- What should I do when the data are Negatively skewed and also have Zero values? 2- what range is defined as corrected skew? when skewness of about 3 tranfered to 1.3 is it considered as a corrected skew or I may need to use inverse for instance? 3- is the range of accepted skewness differ in the fields of studies or it is pure stat? Thank you
Thanks a lot. It was very helpful and you explained very clear. I have a set of skewed data. Please could you explain when we use 'Square Root Transformation'? how could I know which one is more suitable for my data set? Thank you in advance
Hi, excellent video and I have a question my data is positive skewed the skewness is 2.629 but the minimum value is -0.700 .So what can I do for this data to transform it by using Box-Cox transformation
Thank you very much for the video. What non-parametric SPSS based test would you recommend as a substitute for repeated measures ANOVA for a 2x3x2x2 design?
Thanks for the video. Very instructive. I am currently dealing with data that is negatively skewed and furthermore, all values are somewhere within -0.204 and -0.94. To be able to transform these data, do I first reflect and then add a constant? Or what steps are necessary for me to run parametric tests?
Thank you for this video! I transform my data in log10, im doing a regression analysis on money spend, distance, delay etc... im wondering after my regression analysis should i transform those data again for interpretation or im simply analysing the data in "log10"?
Thank you very much for such a quick reply! Before I go into bootstrapping, I was just told that I can/need to run my log-transform over MEAN data per participant per condition (not the RAW data how I was doing before). And actually, when I log-transform my mean data, then residuals are normally distributed now. Does it make sense to you? Is it OK to do so? I hope I was clear.
I found the answer on my question below, but am not sure how this computed variable would look in the numeric expression section. Can I perform all those steps in one new variable? Or do I need to do things separately?
Am I right when I say that you can not just conclude by visually looking at the histogram that you can now run parametric tests on the data? Wouldn't you have to do a test for normality first like the Kolmogorov-Smirnov or the Shapiro-Wilk?
Excellent video! I got a question to ask. In the video you are using skew and kurtosis to analysis the normality, but you did not mention k-s test and shapiro test, i am confused on which method of test of normality should i mainly rely on. I have a set of data and the k-s test is 0.048, which suggests that it is not normal, after transformation, it is still not greater than 0.05, but the graph looks like normal. Can i say it is normal or should i go onto non-parametric? Thank you!
+Ivy Chocoholic The k-s and Shapiro test tend to be overly conservative when analyzing large sample sizes (e.g. >50 subjects). I would suggest using skewness, kurtosis and Q-Q plots.
Excellent video.... Very clearly explained. However after Reflection, my negatively skewed data became positively skewed...then WHAT???? Should I take it as it is coz may be the data is going to remain skewed! In that case should I go for parametric tests? Also, could you please share the reference (book, article, thesis or paper) for using log method as well as reflection method? Thanks a lot for this video. I desperately need it for my thesis. Earlier the better :)
+Kshipra Moghe Some data will remain skewed even after transformation. The alternative is to use non-parametric analysis. A reference would be: Pallant (2014) SpSS Survival manual. (4th ed.) McGraw-Hill.
Nice video! I would like to ask, If I am doing a research based on Paired sample T-test and it is involving vast data sets, should I transform all of those into normal or should I use Non parametric test instead? Considering the data in every sets is consisting approximately 90 data points.
1. How would i do a back-transformation as well as back reflection for your example LG10(91-speed). Would it be Exp(91+speed)? Because this did not work in my data. If i do the transformation and reflection seperately i could easily backtransform, but the reverse reflection back into the original data did not work, except using the options: Transform, recode into an other variable, old and new values. But in a metric variable with many numbers this is exhausting. Is there an other option or an syntax expression for this? 2. What numeric expression in SPSS would i need for the backtransformation of the square-root-function? Can`t find it in SPSS. Could you do an example for srqt reflection, transformation and backwards transformation, reflection?
Great video! I have a question... if your data is negatively skewed and contains 0 should you reflect and add a constant (e.g. +1) or just reflect?? Thanks so much.
Even if you correct for Skewness, under the Kolmogorov-Smirnova test the data was probably not normal. Could we really still use this data using parametric analysis???
could we use any constant value while using log transformation ?! .. for example I have negative skewed so I did use the flowing formula ( log10(2.5- var) ) . which var is the variable that need to transform. I did try many values till I get this one in th end to remove the skewness from the variable. could we do such thing ?!!?!?
Thanks for this tutorial it was a life saver. However, one of my variables when I try lg10(variable+1) as it has 0 in it, when I try and p-p plot error 401# comes up?It says that this variable cannot be log transformed because of the 0 (even though I have added the +1). Is there any solutions you can suggests? Thanks!
If you have negative numbers, the addition of +1 may have still left you with some data points at or below zero. You may have to add a larger constant than 1.
Hi, in a set of data I transformed a skewed data to a normal one. After that I need to do regression analysis, and my confusion is this transformed data is smaller than any other normal data. So which data should I put in? the original one (skewed one) or the normal one? Many thanks!
Pls Help. Two questions. Done transformation using sqrt. Had negative Skewness. After the transformation how do I reflect my result back? (Used no. 8 to reflect my data since 7 was the maximum number. My new square root number are obviously smaller (highest is 2.65) should I use number 3 to reflect them back? 2- If you doing SEM, and you needed to do square root for 5 out of 9 total constructs in model, do you need to square root the rest of variable? will SEM work that way?
Hi the data set i'm working with are percentage scores (on a manual dexterity test) so interval data. The skewness statistic of my data set is -2.351 and std error is 0.255, so i assume I need to normalise my data set by reflecting it then log transforming it. However as my lowest z-score is -4.37, surely when I add a constant it needs to by +4.37 ?
After you reflect it and log10 to transform, how do you back transform it to report the stats? I would appreciate your help greatly as I'm having so much problem with SPSS and this whole stats and my final paper is due soon *cries* please help!!:(
Thank you sooooooo much for this video,..that was such a helpful thing to me...it would be great if you can send to me the page of the book in which is mentionned log transformation when data is zero or negative minutes starting from 8:43 and forward,because the book inquestion is available in my country.i have to use the information above mentionned in my dissertation.sincerly
Hi, I have tried using log transformations to correct normalize skewed data, but the histogram is still skewed, is it possible to repeat the log transformation until the data is more normalized?
Please I am trying to transform my data with skewness of 3.05 but it seems the log transformation is not working. I notice I have a lot of missing values. Could this be the issue?
I have been following the instructions step by step to normalise my data but it will not work with one of my variables, I had two variables that were not normal, one positively skewed and one negatively, the one which is negative even after log transformation remains skewed, and in fact goes even more negatively skewed. I am not sure what I am doing wrong I have done the log transformation numerous times but it still remains skewed. Please can you advise?
I am very grateful you make these videos, you are a real savior.
Could you please give a reference for rule 1 and rule 2 of normality please, I need to mention that in my writing?
And more importantly, How we interpret the result of the transformed data??? it looks complicated, especially when we use log transformation. Many Thanks.
Thank you this has been very helpful you explained better than it was explained in the lecture.
Really good video, the expanations used throughout are excellent
Thank you sooo much for this video. It clarified most of my questions. Thanx a million.
Bloody brilliant! Thank you so much! I am never using Minitab again.
Great video. Clearly explains the process.
Thanks a lot for your prompt reply. Please can you help me on the following questions as well?
1- What should I do when the data are Negatively skewed and also have Zero values?
2- what range is defined as corrected skew? when skewness of about 3 tranfered to 1.3 is it considered as a corrected skew or I may need to use inverse for instance?
3- is the range of accepted skewness differ in the fields of studies or it is pure stat?
Thank you
Thanks a lot. It was very helpful and you explained very clear. I have a set of skewed data. Please could you explain when we use 'Square Root Transformation'? how could I know which one is more suitable for my data set?
Thank you in advance
Hi, excellent video and I have a question my data is positive skewed the skewness is 2.629 but the minimum value is -0.700 .So what can I do for this data to transform it by using Box-Cox transformation
Thank you for this great video. I knew I was missing one click of the mouse somewhere! Really grateful to you! Good karma going your way bud!
Thank you very much for the video. What non-parametric SPSS based test would you recommend as a substitute for repeated measures ANOVA for a 2x3x2x2 design?
Thanks for the video. Very instructive. I am currently dealing with data that is negatively skewed and furthermore, all values are somewhere within -0.204 and -0.94. To be able to transform these data, do I first reflect and then add a constant? Or what steps are necessary for me to run parametric tests?
Thank you for this video! I transform my data in log10, im doing a regression analysis on money spend, distance, delay etc... im wondering after my regression analysis should i transform those data again for interpretation or im simply analysing the data in "log10"?
Thank you very much for such a quick reply! Before I go into bootstrapping, I was just told that I can/need to run my log-transform over MEAN data per participant per condition (not the RAW data how I was doing before). And actually, when I log-transform my mean data, then residuals are normally distributed now. Does it make sense to you? Is it OK to do so? I hope I was clear.
I need to replace the missing data first and then deal with skewedness, am I right? Thanks.
I found the answer on my question below, but am not sure how this computed variable would look in the numeric expression section. Can I perform all those steps in one new variable? Or do I need to do things separately?
Am I right when I say that you can not just conclude by visually looking at the histogram that you can now run parametric tests on the data? Wouldn't you have to do a test for normality first like the Kolmogorov-Smirnov or the Shapiro-Wilk?
Excellent video! I got a question to ask.
In the video you are using skew and kurtosis to analysis the normality, but you did not mention k-s test and shapiro test, i am confused on which method of test of normality should i mainly rely on.
I have a set of data and the k-s test is 0.048, which suggests that it is not normal, after transformation, it is still not greater than 0.05, but the graph looks like normal. Can i say it is normal or should i go onto non-parametric?
Thank you!
+Ivy Chocoholic The k-s and Shapiro test tend to be overly conservative when analyzing large sample sizes (e.g. >50 subjects). I would suggest using skewness, kurtosis and Q-Q plots.
Why do you need to transform the data? Is this approach better than using non-parametric tests?
Thank you.
Excellent video.... Very clearly explained. However after Reflection, my negatively skewed data became positively skewed...then WHAT???? Should I take it as it is coz may be the data is going to remain skewed! In that case should I go for parametric tests? Also, could you please share the reference (book, article, thesis or paper) for using log method as well as reflection method? Thanks a lot for this video. I desperately need it for my thesis. Earlier the better :)
+Kshipra Moghe Some data will remain skewed even after transformation. The alternative is to use non-parametric analysis. A reference would be: Pallant (2014) SpSS Survival manual. (4th ed.) McGraw-Hill.
Nice video! I would like to ask, If I am doing a research based on Paired sample T-test and it is involving vast data sets, should I transform all of those into normal or should I use Non parametric test instead? Considering the data in every sets is consisting approximately 90 data points.
+michael fonglius If you can transform, that would be my first choice. If transformation does not work, then use non-parametric analysis.
1. How would i do a back-transformation as well as back reflection for your example LG10(91-speed). Would it be Exp(91+speed)? Because this did not work in my data. If i do the transformation and reflection seperately i could easily backtransform, but the reverse reflection back into the original data did not work, except using the options: Transform, recode into an other variable, old and new values. But in a metric variable with many numbers this is exhausting. Is there an other option or an syntax expression for this?
2. What numeric expression in SPSS would i need for the backtransformation of the square-root-function? Can`t find it in SPSS. Could you do an example for srqt reflection, transformation and backwards transformation, reflection?
Anna ST !. Not that I am aware of.2. You would perform the opposite of the function you used to transform the data.
Great video! I have a question... if your data is negatively skewed and contains 0 should you reflect and add a constant (e.g. +1) or just reflect?? Thanks so much.
You should be able to reflect only.
*****
Thanks. And how to reflect the transformation back for data analysis?
DinosaurRock100 This video should help: ua-cam.com/video/P4IVjMEa6bM/v-deo.html
Even if you correct for Skewness, under the Kolmogorov-Smirnova test the data was probably not normal. Could we really still use this data using parametric analysis???
could we use any constant value while using log transformation ?! .. for example I have negative skewed so I did use the flowing formula ( log10(2.5- var) ) . which var is the variable that need to transform. I did try many values till I get this one in th end to remove the skewness from the variable. could we do such thing ?!!?!?
Yes, you can use any number as a constant.
Thanks for this tutorial it was a life saver. However, one of my variables when I try lg10(variable+1) as it has 0 in it, when I try and p-p plot error 401# comes up?It says that this variable cannot be log transformed because of the 0 (even though I have added the +1). Is there any solutions you can suggests? Thanks!
If you have negative numbers, the addition of +1 may have still left you with some data points at or below zero. You may have to add a larger constant than 1.
Thanks man! It's really much helps
Hi, in a set of data I transformed a skewed data to a normal one. After that I need to do regression analysis, and my confusion is this transformed data is smaller than any other normal data. So which data should I put in? the original one (skewed one) or the normal one? Many thanks!
and can I put this transformed data and other originally normal data together in a factor analysis and regression analysis?
+Jane Wei Since regression assumes data is normally distributed, I would use the normal data. It can be used in factor analysis.
Pls Help. Two questions. Done transformation using sqrt. Had negative Skewness. After the transformation how do I reflect my result back? (Used no. 8 to reflect my data since 7 was the maximum number. My new square root number are obviously smaller (highest is 2.65) should I use number 3 to reflect them back?
2- If you doing SEM, and you needed to do square root for 5 out of 9 total constructs in model, do you need to square root the rest of variable? will SEM work that way?
Hi the data set i'm working with are percentage scores (on a manual dexterity test) so interval data. The skewness statistic of my data set is -2.351 and std error is 0.255, so i assume I need to normalise my data set by reflecting it then log transforming it. However as my lowest z-score is -4.37, surely when I add a constant it needs to by +4.37 ?
Zizi Lee Yes, that is correct.
After you reflect it and log10 to transform, how do you back transform it to report the stats? I would appreciate your help greatly as I'm having so much problem with SPSS and this whole stats and my final paper is due soon *cries* please help!!:(
This video should help: How to Use SPSS: Reporting Log Transformed Data
Thank you:)
Thank you sooooooo much for this video,..that was such a helpful thing to me...it would be great if you can send to me the page of the book in which is mentionned log transformation when data is zero or negative minutes starting from 8:43 and forward,because the book inquestion is available in my country.i have to use the information above mentionned in my dissertation.sincerly
Could you let me know which book you are referring to?
Mertler and Vannatta, Advanced and Multivariate Statistical Methods(2010), Pyrczak publishing
114kha pp. 32-34
You are the best!
Hi, I have tried using log transformations to correct normalize skewed data, but the histogram is still skewed, is it possible to repeat the log transformation until the data is more normalized?
No, if it is still skewed after log transformation you should consider non-parametric data techniques.
Please I am trying to transform my data with skewness of 3.05 but it seems the log transformation is not working. I notice I have a lot of missing values. Could this be the issue?
+AYODEJI IYANDA It may be due to the presence of outliers.
I have been following the instructions step by step to normalise my data but it will not work with one of my variables, I had two variables that were not normal, one positively skewed and one negatively, the one which is negative even after log transformation remains skewed, and in fact goes even more negatively skewed. I am not sure what I am doing wrong I have done the log transformation numerous times but it still remains skewed. Please can you advise?
Thank you very much!
Thank you very useful
thank you!