How to create index using Principal component analysis (PCA) in Stata

[@polinakoriukina699]

Thank you so much, I couldn't figure it out for hours until I found this video. Literally a lifesaver

[@Munyaldhinoooo]

Thank you!!! You have simplified this process so well and understandable even by non-experts!!

[@Khadija651]

JazakAllah khairan kaseera brother.... This helped me alot

[@khadimhussainmalik3284]

Dear Sohaib Ameer. Being a professional we must follow the ethics and rules of every profession. You just run the pca command and predict the index without going in details analysis of the reslults. WE must ensure the credibility and reliability of the index that is going to represent an issue in the research paper. As I know the process of PCA is not much easy as you explained here. There are lots of things that being a researcher we must ensure.

[@mohdkashif86]

Dear @khadim you should create a video to cover up all those " lots of things that being a researcher must ensure." Wasaalam

[@PragatiCreationsRnD]

Easy and well explained. Thanks

[@frmabhijit]

All variables are having positive values. But, your FD index is having both positive and negative values. As I understand, the index should not have negative values. How to interpret the negative index values?

[@lutfaferdous8236]

Many thanks Sohaib. It was a great video

[@sherzadshahab9628]

Thanks a lot. It is very useful.

[@freemanNsoh]

Thank you so much for the video....🙏

[@bakytzholamanova2560]

Thank you!! Your video helped a lot!

[@skmosharafhossain1192]

Many thanks for this video. Very helpful. I want to understand a basic thing. Why is the weighted index value (here the fd) more than 1? It should ideally be 0-1. Can you please clear my doubt.

[@HCN_PROFESSOR]

well explained bro

[@Mat-mt8pk]

Thank you. But I see one mistake. You found that it makes sense to take two components of the pca operation since it was showing an eigenvalue greater than 1 for the first 2 components, then why did you predict only the first component. This will lose a lot of information. You should've used: predict pc1 pc2

[@oforicharles2281]

what happens after predicting pca1 and pca2. You’ll obtain two variables. Is the sum of the two going to be your index now? Kindly advise!

[@Chamomile136]

how can i predict pc1 pc2? what command should i use? would you please kindly tell me how i can get the result of the index after predict those 2 pc?

[@johnnyzhao1655]

Why you have to rotate the original PCA?

[@kibromadinoabate4777]

It is more informative video, thank you so much for that. How about if we have binary response for items (like tv) and combination with continuous variable (number of bicycle)? Can we use pca or polychoric pca? One more, what is the implication of negative fd value in you output?

[@adnansabzi]

thank you sir really more informative, sir if its possible plz upload a video about extreme bound analysis

[@SONALISINGH-dj6jg]

please tell which variables you have used to make FD

[@kibromadinoabate4777]

Thank you so much. Would you help us know the questionnaire for regular PCA and polychoric PCA? Is there difference among them on variable indicators ?

[@nomanrasheed9644]

Do stata standardized variables automatically?as you haven't standardized the variables. variables with different scales need standardization.

[@yapshen2076]

Thank you very useful. I want to construct an index using two-step PCA. This is the first step. Is it possible to shed light on 2 things: 1. How to carry out 2nd step when I get the results (prediction) from each dimension? 2. How to convert the result (prediction) to an index between 0 and 1

[@khalilrahman2339]

Sir if we have some variable which is negatively related and some have positive relationship, so what should we do in that case. In normalization we have two different formulas for negative and positive relationship with defended variable for index making through normalization and aggrigation, but my question is what should we do if we apply PCA

[@Jatindersingh-wo5hf]

You should clarify the basics like why PCA on which type of variables PCA is applicable. Hiw to scale different variables etc.

[@Chamomile136]

Please, did you ever get an answer to this question?

[@Jatindersingh-wo5hf]

Never

[@Chamomile136]

@@Jatindersingh-wo5hf in terms of scaling, when should i use standardization or normalization? can you kindly tell me about this

[@Jatindersingh-wo5hf]

@@Chamomile136 see these are different type of tools to make data unit free.so u can use both

[@Chamomile136]

@@Jatindersingh-wo5hf thank you, i want to ask one more question. If i have 2 variables, one is better if the number is greater and the another is worse is the number is greater, then how can i rescale this?

[@ArpitBarman-g6s]

hey thanks for the useful video. I have just one query, Do we need to standardize the variables when applying PCA command in STATA 17 or applying the PCA command will work only? will it automatically standardize the data when PCA cmd is applied? Thanks

[@tubihemukamamethodius6952]

The command standardises data automatically

[@Chamomile136]

did you find the answer? can you please kindly tell me

[@aaizazpatel4626]

Can you please explain why did you rotate the values?

[@PeaceinPod]

Sir how to find out the concentration index of inequality using STATA?

[@esanmathew-c8s]

Hello , I need information on how to calculate Livelihood Vulnerability Index. Thanks

[@gayatrigogoi571]

can you please say which variables did you take

[@alinik4737]

How to treat messing values? I watched your video in spss. but stata menus dont have those options in spss. please answer Im in trouble.

[@Garrettthethief]

Hi there. You forgot about saying, that the predict is only accurate for this model used. Do you know an alternative for the predict command via mean index as well?

[@naallahbiliqees7957]

Well done sir, please i want to compute index also but some observations are missing, what do i do. Secondly, in one of the dimensions only one indicator has variable other do not, please how can i calculate the index

[@sohaibameer3742]

Naallah Biliqees kzbin.info/www/bejne/npbZmGNsm7Wsfas follow this video for missing data case and the rest. thank you!

[@smeetb01]

बहुत अच्छा शुक्रिया!

[@makitsion8652]

hi my reasearch title is the role of credit on women economic empowerment and can i use tobit model how can i use please show me

[@mohdkashif86]

THANK YOU Ameer S.

[@n_a05]

how we can estimate the scores of these variables?

[@efclientopsurvey3868]

Shouldn't the base year be equal to 100 (or 1) for an index?

[@thesuprati]

Why did you use the rotate command.......You did not give reasons for rotating. in my view if you have variables which are correlated and which might create a problem of multicollinearity in your model, then you use PCA to get rid of this problem. Would you please give some specific statistical justification of using the rotate PCA command. thanks

[@hassaanahmad4770]

same question.... one more question..... after analysis you got 2 components but you predict only one i cant understand this.....can we use only one whether there will be 2 or 3 components.....??? please explain

[@juheesingh1157]

To my understanding Rotation is used when either variables have same weight or all variables attach to one factor only .

[@juheesingh1157]

@@hassaanahmad4770 he got two factors on the basis of which he created a single index.

[@Postdoclife123]

for rotation in PCA questions - transformation to convert a set of values of possibly M correlated variables into a set of K uncorrelated variables called principal components. Varimax rotation-It changes the coordinates that maximize the sum of square loading.

[@nsreenali1847]

@@juheesingh1157 do you have idea, how i can use PCA with variables that have different weighted

[@omersahil339]

hi, i just saw your video, not related to my problem. I want make the index of 5 different variables on the basis of decile rank. your cooperation considered highly appreciated in this regard.. thanks

[@danielkrupah]

I need this too. Please were you able to do it? [e-mail]

[@rishikanayyar1630]

Can we do PCA for all the cross-sections (i.e. Pakistan, India etc) together?

[@sureyyaylmaz6616]

Did you figure out how do you PCA for cross-sections?

[@rashmisajwan1724]

Thank you for the video, Sir I would like to know that how do I use PCA on panel data. I have data of "crime against women" which is cross sectional and time series. And I want to create an index using PCA. But how should I approach pca for panel data in stata . Please explain. Or if possible please make a video.

[@parfaitberi3974]

Please, did you ever get an answer to this question?

[@mamadouthiam7575]

Bonjour. Est ce que vous avez la réponse à cette question ?

[@humakhalid6038]

Helps a lot

[@sidaliseghiri1655]

Thank you so much

[@JMRG2992]

How can I make the Index lie between 0 and 1 ?

[@manish9555]

Use Normalised values then you will be able to get index value between 0 and 1.

[@JMRG2992]

@@manish9555 Well, standard deviation did not work. But, using a rescaling process of the index got the job done.

[@clarajoanjoachim2179]

@@JMRG2992 hi, can you please advise with this?

[@JMRG2992]

@@clarajoanjoachim2179 Once the PCA is estimated, you can compute the first principal component of the PCA. with that, you can rescale the principal component into an index between 0 and 1. The process, implies the formula: Xi - X max / Xmax - Xmin

[@clarajoanjoachim2179]

@@JMRG2992 thanks, I was thinking if to rescale the variables first (using the formula you gave) and only apply PCA. does it makes any difference?

[@joselinceron8602]

What happens when i want 3 pc for my ibdex?

[@madlipzarabic2298]

please show the scale of the variable representing financial development, i want an answer please..