How to Identify and Treat Outliers in Stata | Stata Tutorial

[@Mat-mt8pk]

Methods of finding outliers 1:14 #1. Sorting 2:52 #2. Box Plot 6:04 #3. Extremes 10:05 #4. Histogram 10:50 #5. Spike Plot 11:42 #6. Zscore Treatment 13:07 #1. Keep outliers 13:42 #2. Correct error 14:23 #3. Winsorization 19:06 #4. Trimming

[@thedatahall]

Thanks for the efforts

[@gonout8402]

You have explained everything that my professor taught me in 2 months in just 20 minutes and it's is much more understandable and useful. Thank you very much

[@thedatahall]

😄

[@rouniktalukdar872]

Amongst the nicest video lecture that I have come across on this topic.. Thanks a lot. please keep uploading more contents on STATA.

[@thedatahall]

Thanks for the appreciation

[@wilsonahinful5127]

This is all that I have been looking for, thanks very much indeed

[@addisugetahun1441]

Thank you for your nice and clear lecture in identifying and treating outliers.

[@alphadie2012]

Clear and concise explanation. Thank you

[@tomaxow]

Really well done and explained

[@thedatahall]

Thanks

[@jemalhassen2841]

It a very helpful video. Thank you!

[@thedatahall]

Thanks. Keep sharing

[@jibrilyero2263]

Great job 🎉

[@danishjunaid1659]

Very well explained

[@thedatahall]

Thanks

[@isaacasante4060]

Awesome video. Could you please do a similar one using panel data.

[@thedatahall]

Sure will make a video on that

[@shafiqullahyousafzai15]

Thanks from Afghanistan

[@korneliuslanggason5477]

thank you for the explanation.

[@thedatahall]

Thanks

[@yilebesaddisu5314]

Thank you dear, very helpful!!

[@lottet1945]

Thank you for this clear explanation! Do you have a video on Cook's distance and Mahalanobis distance in Stata by any chance?

[@thedatahall]

Thanks for watching the video. Unfortunately i currently dont have video on this. I will see if in future i might add this. But if u r interested in spss then there are videos on KZbin

[@aibannongspung1765]

Thank you so much for this insightful video !! Suppose I want to trim the top and bottom 0.1 % of the distribution .How do I write the command ?

[@thedatahall]

I have never tried with decimals but the command will look like winsor2 variablename, trim cut(0.1 99.9)

[@thedatahall]

Let me know if it works

[@RafiaAli-n8e]

Hi, hope you are doing great. Can you share the link of multivariate outliers, I am not able to find it?

[@thedatahall]

Thanks for your kind words. Unfortunately we haven't made any video on multivariate outliers. I will add that in my todo list

[@RafiaAli-n8e]

It would be highly appreciated.@@thedatahall

[@AhaNYS]

Thank you for the video! I have a question, I want to use ssc extremes among subcategories. How can I apply this extremes for every subcategory??

[@thedatahall]

U can try bys category: extremes etc etc

[@badiahahmed2085]

Thank you for your great video. I have a question please, After using the Winsorization, can I take the logarithm for some variables? Thank you.

[@thedatahall]

Yes you can take log after winsorization. But be advised that after taking log the interpretation of coefficient changes to percent change. I am soon going to make a video on functional forms, so if u dont have the idea on interpretation after taking log then that video will help.

[@badiahahmed2085]

@@thedatahall Thank you for your response, that will be great. MANY THANKS

[@tranglephuong1896]

Can you give me the dataset you run in video?

[@thedatahall]

unfortunately i have misplaced the data and do file for this specific video.

[@shrinjoy1234]

How do we use winsor command if we want to replace outliers with Q3+1.5 IQR Can we use winsor command to handle outliers of multiple columns in one go? Please advise.

[@thedatahall]

it is not possible using winsor or winsor2 command. you will have to write code for it. one way is to create a variable that will store the value of Q3+1.5iqr and then u can use that to replace in your main variable

[@atiyaabdulkarim716]

A quick question, if we use sort function, will it allign all other observations in other variables? For eg. If we Sort by price, but we have other variables on age education and i.d. No. So after sorting by price, would it keep track of age and education with respect to i.d. after sorting or only one variable would be sorted not others, this can create problems, No?

[@thedatahall]

In stata the sort comment will keep tract of all variables and sort them simultaneously. The whole row will move and not the specific column of price.

[@thedatahall]

Sort only sorts in accending order, there is another command gsort -price so now it sort in descending

[@atiyaabdulkarim716]

Can you tell us/take us through calculator functions in stata (syntax for exponent and complex function)

[@thedatahall]

Sure, u want me to make a video on arithmetic etc functions in stata?

[@atiyaabdulkarim716]

@@thedatahallthank you for getting back to me. I am a medical student and i have to use calculate function in stata to generate a new variable. My problem is that some components are used in exponent form, if you look at MDRD equation to define chronic kidney disease or CKD EPI equation, you will see serum creatinine levels, age are entered in the formula. My specific question is if i want to use this information from some variables in my data set, how can i do this. I tried exponent function but my calculations appear to be incorrect and it seems i am not following the right steps. I would highly appreciate if you could make a video or may be if you can give me a feedback.

[@thedatahall]

What command did u used, if u used exp() function then thats to invert log... If u email me the equation at [e-mail] and might be some sample data or the command u have used i will look into it. If u wanted to take power e.g. square of a number then u do gen newvariable=oldvariable^2

[@thedatahall]

I searched for mdrd equation but i am not sure i found the right one

[@atiyaabdulkarim716]

@@thedatahall thank you for getting back to me, here is the link: patient.info/doctor/estimated-glomerular-filtration-rate-gfr-calculator Normal creatinine values range between 0.6 to 1.2 mg/dl...so one can use values at higher end or perhaps old age and see what is the filteration rate....

[@alfinasintiya7477]

saya tidak dapat menggunkan "extremes" adakah solusinya?

[@thedatahall]

i just used extreme, its working fine with me what error u are getting? saya hanya menggunakan "extremes", berfungsi dengan baik dengan apa ralat yang anda dapat?