Its upto you if you want to load them in one index or make multiple indices, For this you need to study the concept of discriminant and convergent validity. If all the items are highly correlated with each other loading them in one index is better as it would remove multicollinearity
@@nomanarshed Sir forexample like making 1 index with economic variables and then other index with social variables and then applying regression with them on one variable is that possible
since all of your data is in same scale (binary) you can average them to form an index unless you want unequal weights in that case you can use factor analysis.
Yes it is providing us a linear combination of the data measing latent variable in continuous form. If its values cannot be than zero then the error term of regression should also be not zero as it is also an index as a resultant of linear combination of dv and ivs
Thankyou so much.
Welcome!
please how can we obtain the contribution or weights of each variable in the index
See factor loadings generated after factor analysis
If we want to construct weighted index we choose symetric weight or loading weight please help me
It depends on theory and requirement. For symmetric weights ideally the data has same units.
Thanks for this very informative video
Glad it was helpful!
if we have many variables under different cateogories can we do pca for each category and then run the panel regression?
Its upto you if you want to load them in one index or make multiple indices, For this you need to study the concept of discriminant and convergent validity. If all the items are highly correlated with each other loading them in one index is better as it would remove multicollinearity
@@nomanarshed Sir forexample like making 1 index with economic variables and then other index with social variables and then applying regression with them on one variable is that possible
Yes then these 2 index variables can be used in regression as IVs only if they are not highly correlated with each other.
@@nomanarshed thankyou sir
Respected Sir
How we can create index for variables on binary scale
since all of your data is in same scale (binary) you can average them to form an index unless you want unequal weights in that case you can use factor analysis.
You might be a great statistician, but not at all even a basic level teacher. !
Yes this is a research related channel. I am learning and your guidance will be helpful.
If the index results are less than 0 How I could deal with it?
Index is a normalized data where zero means mean value and negative values means that the paticular observation is less than mean.
@@nomanarshed as far I know the index should be between 0-1 or from 1-100 right?
But the PCA results are not on this ranges what should I do?
It is a normalized variable not a index which ranges from - infinity to infinity
@@nomanarshed so I can use it even if it is less than 0
Yes it is providing us a linear combination of the data measing latent variable in continuous form. If its values cannot be than zero then the error term of regression should also be not zero as it is also an index as a resultant of linear combination of dv and ivs