Thanks for this comprehensive video. The last regression (after the imputation was done) gives results based on the exact (full) number of observations. But once I save my dataset and then reopen to run regressions, it gives me results where the number of the observations now exceed the regular size of the sample. How is this possible? Where did I make mistake? Many thanks
If attack is the dependent variable, I don't it's the appropriate variable to be used for imputing missing values of the predictor variable bmi. If you run a logistic regression with a dichotomous attack and continuous bmi, wouldn't the association between the 2 variables be overestimated? I would probably select a covariate/s that is/are highly correlated with the variable being imputed but then I'll have collinearity problem. Any guidelines on choosing the variables that should be used for imputation?
How to do MI for panel data with multiple variables (N=30 countries, T = 21 years), some countries have missing values for some variables, please someone guide me, thank you in advance
thank you ! 1/ how to do it with 2 (or more) variables where we observe missing data ? (in your exameple it is bmi only. how to do it if it would be 'attack' and 'bmi' that you observed missing data ?) . 2/ how to report the statistical analysis part in an article ? thank you for your response
For more than one variables just include the other variable you need to impute when using the impute command, make sure you also use register command on that variable
Why you performed the logistic regression model from the multiple imputation panel? It would be the same if you had launched the logistic regression model from the logistic regression command?
+Gemari Ortega - mi set - creates a new variable named "_mi_m". Each imputed dataset is numbered 0 through M where M is the number of imputed datasets. So if you created 5 imputed datasets, the variable "_mi_m" would contain 0-5. The observations that correspond to _mi_m=0 are the original data. You can use _mi_m to check each imputed dataset separately. If you would like more information, please contact Technical Services.
i have about thirty variables, 700 units of analysis and 10 years worth of data and have not been able to find anything there that would help me to do multiple imputation with this dataset, maybe you could direct me to a specific page if you know there might be something useful?
@@RPDBY Hope you are well and expert in STATA. Do you mind sharing your experience with me please? After 4 years, still there is no helpful guidance on how to do it for unbalanced panel data of multiple variables.
@@vitaltopics316 still no idea, but after doing some research I concluded it is not worth doing it because it does not help much to improve the analysis quality (imputations will always be questioned by any serious academic) but is a lot hassle. either get more data or work with what you have
The command shown in the title screen is "mi regress". Note that the actual command is "mi impute regress".
Thanks for this comprehensive video. The last regression (after the imputation was done) gives results based on the exact (full) number of observations. But once I save my dataset and then reopen to run regressions, it gives me results where the number of the observations now exceed the regular size of the sample. How is this possible? Where did I make mistake? Many thanks
If attack is the dependent variable, I don't it's the appropriate variable to be used for imputing missing values of the predictor variable bmi. If you run a logistic regression with a dichotomous attack and continuous bmi, wouldn't the association between the 2 variables be overestimated? I would probably select a covariate/s that is/are highly correlated with the variable being imputed but then I'll have collinearity problem. Any guidelines on choosing the variables that should be used for imputation?
Very helpful, even for advanced users of Stata. Thanks!!!
Very helpful.
What about imputing categorical variables, the video can be just as brief?
Can i impute a dependent variable? Thank you!
How to do MI for panel data with multiple variables (N=30 countries, T = 21 years), some countries have missing values for some variables, please someone guide me, thank you in advance
good tutorial,how can i impute categorical variables to set for survival analysis?
thank you ! 1/ how to do it with 2 (or more) variables where we observe missing data ? (in your exameple it is bmi only. how to do it if it would be 'attack' and 'bmi' that you observed missing data ?) . 2/ how to report the statistical analysis part in an article ? thank you for your response
For more than one variables just include the other variable you need to impute when using the impute command, make sure you also use register command on that variable
If i wanted to impute data for a PCA, how would i do this? Simply impute without specifying a regression model?
Why you performed the logistic regression model from the multiple imputation panel?
It would be the same if you had launched the logistic regression model from the logistic regression command?
Hi Thanks for this video. In the beginning of the video you refer to a manual. Can you please provide a link for that manual ?
www.stata.com/bookstore/multiple-imputation-reference-manual/
A very helpful video! Thank you so much!
Chuck. thanks very much for the information. How can I impute more than 1 variable at the same time?
Nitin Rajput you can use MICE (multiple imputation using chained equations)
est how to check if i should go for fixed or random model post-imputation, hausam dont work on weighted data
Thanks!
Hi thanks for the video, another one about multivariate imputation would be useful. Thank you.
I'm sorry, what manual is it that you are referring to?
Stata's Multiple Imputation Reference Manual.
okie now i imputed and wanna try panel model, but hausman wont work , help plz
How do I check the imputed data?
+Gemari Ortega - mi set - creates a new variable named "_mi_m". Each imputed dataset is numbered 0 through M where M is the number of imputed datasets. So if you created 5 imputed datasets, the variable "_mi_m" would contain 0-5. The observations that correspond to _mi_m=0 are the original data. You can use _mi_m to check each imputed dataset separately. If you would like more information, please contact Technical Services.
The screen is hard to see.
+Soomi Lee HI! Be sure to set your video playback to the highest quality. The HD setting is preferable.
useless tutorial. how to do it with panel dataset and multiple variables?
RPDBY you should read the manual
i have about thirty variables, 700 units of analysis and 10 years worth of data and have not been able to find anything there that would help me to do multiple imputation with this dataset, maybe you could direct me to a specific page if you know there might be something useful?
@@RPDBY Hope you are well and expert in STATA. Do you mind sharing your experience with me please? After 4 years, still there is no helpful guidance on how to do it for unbalanced panel data of multiple variables.
@@vitaltopics316 still no idea, but after doing some research I concluded it is not worth doing it because it does not help much to improve the analysis quality (imputations will always be questioned by any serious academic) but is a lot hassle. either get more data or work with what you have
@@RPDBY Thank you so much for the fastest reply. I will work with the data I have. You saved me a lot of time. THANK YOU SO MUCH
stop being so boring