Converting Non Normal Distribution / Data TO Normal Distribution Data | Using Minitab & Excel

[@abrarmujeeb3695]

Excellent video

[@AbrarKnowledge]

Thanks lot

[@shivashanker9581]

Great video Sir. Thanks for making

[@AbrarKnowledge]

Thanks a lot

[@tknithi]

Really usefull video. Thank you.

[@AbrarKnowledge]

Thanks a lot

[@sanjaydesai8263]

Very useful, nicely demonstrated

[@AbrarKnowledge]

Thanks a lot

[@bandungmee]

Sir Thank you for your video. I have two questions: 1. When we do the conversion, is the variable using compute sum data or the responses from the respondent? For example, I have variable and sub-variable 2. If the data is normally distributed after conversion, that means we are using the converted data for further analysis i.e for SEM? Thank you

[@AbrarKnowledge]

Thanks for the questions and let me elaborate a bit

[@AbrarKnowledge]

1. If you are referring to 2 variables then must be looked for normality test and decide further 2 yes but this is tricky because lamda value doesn't bring an actual mean value.. so it depends

[@bandungmee]

​@@AbrarKnowledge thank you for your quick reply, sir If I can explain further: a. my model is a third-level higher-order construct. i.e independent variable --> sub-variable -> indicator. Tested using KS, and all variables are not normally distributed. I am currently converting to lg10, sqrt and inverse to check the normality and using computed sum of variables. b. While testing the skewness and kurtosis, only a few do not match with the range of -+ 1.00. I also checked based on the other range as some papers mentioned the range -+1.00 may be a bit tough for a social science study. Ranges suggested by some other references: -.249,2.33 (skewness) -1.92, 7.42 (kurtosis) -2.00,2.00 (skewness) -5.00, 5.00 (kurtosis) -3.29,3.29 (skewness) -3.29,3.29(kurtosis) Results from the above testing show: a. Checking with different ranges of skewness kurtosis from different references, shows some variables are normally distributed. b. Checking with lg10 and sqrt; there are some variables that are not normally distributed (some of these variables are normally distributed if I am using a different range of skewness kurtosis). I am a bit confused about how to decide with some of the not-normally distributed variables: a. Assuming that there are still some variables that are not normally distributed, what should I do next with the variables that are not normally distributed? b. Should I remove the variables? or use the lambda value (compute sum value)? b. Do I still need to proceed with a non-parametric test? Thank you Sir.

[@TheGardenHeart]

Sir, I had a doubt, I have two (50group A and 50 group B) study groups, for which I'm measuring the same parameter, should I consider the data set as whole, 100 or 50/50 for the transformation

[@AbrarKnowledge]

Great question...for variable data, Individually you can test the normality (Minitab - Stat - Basic Statistics- Normality test ) or Minitab- Stat - Basic Statistics- Graphical summary...any one of the tests can be done to find p value..if p value is greater than 5% or 0.05 then no transformation is required...if no..then transformation or grouping subgroup is required..Hope it clarifies

[@TheGardenHeart]

Thank you Sir 😊

[@sanjaydesai8263]

Can we get excel template

[@AbrarKnowledge]

Sure, will send in whatsapp

[@Big2010Daddy]

Can someone share the data used in this video please? Thank you.

[@AbrarKnowledge]

Sure we can share...email [e-mail]