Hi Alan, your video is great! Thank you! I am having trouble producing interaction plots after using FIML method treating missing values. I would greatly appreciate your advice on it.
Ekeoma, if you interact two dummy variables you produce four outcome groups. You can use Stata's factor notation to do this in different types of models. See help fvvarlist for detail. Here is a quick (and analytically meaningless) example that creates a dummy variable and then uses it in a regression model interacted with another dummy variable. I have included a covariate as well. The plotted results show the relationship between displacement and fuel efficiency across the four groups defined by the interaction term. Hopefully, this gives you some ideas on how to set up your model. sysuse auto, clear * Create dummy variable for this example sum weight, meanonly gen weightdum=weight>=r(mean) * This is the interaction to be estimated tab weightdum foreign, nolabel * Use regress with factor notation (see help fvvarlist) regress mpg displacement i.weightdum##i.foreign margins weightdum#foreign, at(displacement=(75(25)200)) marginsplot, noci scheme(s2color)
The data are available from different places, but this is a good place to download subsets of the General Social Survey: gssdataexplorer.norc.org/ The margins & marginsplot commands can be used after many estimating routines including regression, logit, & probit models. It is used the same way for each different type of model.
c.educ##c.educ is like generating a quadratic effect? That is, its the same as gen edu2=educ*educ??? You did this because you assume that educ and prestige did not have a linear relationship but a quadratic one? I can you all this margin commands after a logistic regression as well, right? It would be great if you do a video explaining margins (post estimations commands) for logistic regressions. Thank you very much! Marvin
+Marvin Aliaga Yes, that model was testing for a quadratic relationship. It is best to use factor notation so you can use margins and Stata can keep track of all your variables. This is particularly important with quadratics and logistic (or other limited dependent variables). When you use margins after a logistic regression, margins reports probabilities, which are easy to interpret. Then, you can use marginsplot to visualize the model. This is a real strength of Stata. Best, Alan
Hello, I have a question can I use the "margins at" command for values that are not in the sample? Let's say the max. education level in the sample is 10 years and I want to know the potential effect of 15 years, can I use "margins, at (15)" for that ?
Hi Tony, yes, this can be cautiously done, but be careful. Making predictive estimates outside the range of the data used in your model can produce nonsensical results. For example, based on General Social Survey data I can show you a model indicating an average decrease in sexual frequency of 1.2 times per year for each year of aging. This model probably works pretty well near the middle of the distribution of age, about 48 years old. This model includes respondents between the ages of 18 and 89 and may provide some information about octogenarians, but is not useful at all when examining people who are 100 years old! Here is another example that demonstrates that you can use values outside of the range of your data in the margins command, but also produces nonsensical results. I used the auto dataset included with Stata to estimate a fuel efficiency model relating MPG to the weight of cars. Then, I used margins to estimate the average MPG for an M1 Abrams tank that weighs approximately 120,000 lbs., clearly outside the range of my data! Run this little program to see the results. Best, Alan sysuse auto, clear /* Estimate a simple model */ regress mpg c.weight##c.weight /* Get some reasonable bounds for weight */ estat summarize /* Margins within the range of the data */ margins, at(weight=(1800(200)5000)) marginsplot, recast(line) recastci(rarea) scheme(s1mono) /* Margins outside the range using the weight of an M1 Abrams tank */ margins, at(weight=120000)
Hi Jodi, Yes, you can have two or more IV's in your model and can use them to estimate the marginal effects. It requires some thought about what you want to see from your model. I have provided some Stata code below that may give you some ideas about how to do this. sysuse auto, clear regress mpg weight displacement estat summarize margins, at(weight=(1800(200)4800)) marginsplot, recast(line) noci name(m1a, replace) margins, at(displacement=(100(50)400)) marginsplot, recast(line) noci name(m1b, replace) margins, at(weight=(1800(500)4500) displacement=(100(100)400)) marginsplot, recast(line) noci name(m2, replace) regress mpg c.weight##c.displacement margins, at(weight=(1800(500)4500) displacement=(100(100)400)) marginsplot, recast(line) noci name(m3, replace)
"The White Collar Holler" by the great Stan Rogers. Traditional music. My favorite song by Rogers is "Northwest Passage". I think Pierre Trudeau called it the unofficial national anthem of Canada!
![MV วิถีชีวิตใต้ - ผา อัยย๊ะ [Official MV HD]](http://i.ytimg.com/vi/nODlkOUu9B0/mqdefault.jpg)
![⭐ฮิตในTikTok!! ( หมูเด้ง MooDeng ) Ver. แดนซ์โจ๊ะๆ ReMix BY [ ดีเจกิต รีมิกซ์ ]](http://i.ytimg.com/vi/3krsyfWUeHU/mqdefault.jpg)
Hi Alan, your video is great! Thank you! I am having trouble producing interaction plots after using FIML method treating missing values. I would greatly appreciate your advice on it.
so great video!!! very understanding! Greetings from Chile
Fantastic video, much appreciated!
I'm glad you found this video useful.
Wow! More more more! Greetings from Germany
Thank you so much for this video. Although my question is 9 years late how about interacting two dummies
Ekeoma, if you interact two dummy variables you produce four outcome groups. You can use Stata's factor notation to do this in different types of models. See help fvvarlist for detail.
Here is a quick (and analytically meaningless) example that creates a dummy variable and then uses it in a regression model interacted with another dummy variable. I have included a covariate as well. The plotted results show the relationship between displacement and fuel efficiency across the four groups defined by the interaction term.
Hopefully, this gives you some ideas on how to set up your model.
sysuse auto, clear
* Create dummy variable for this example
sum weight, meanonly
gen weightdum=weight>=r(mean)
* This is the interaction to be estimated
tab weightdum foreign, nolabel
* Use regress with factor notation (see help fvvarlist)
regress mpg displacement i.weightdum##i.foreign
margins weightdum#foreign, at(displacement=(75(25)200))
marginsplot, noci scheme(s2color)
Do you possible to have the data you use it? Thanks it’s very important topic and how we can do margins for multivariate probit?
The data are available from different places, but this is a good place to download subsets of the General Social Survey: gssdataexplorer.norc.org/ The margins & marginsplot commands can be used after many estimating routines including regression, logit, & probit models. It is used the same way for each different type of model.
@@smilex3 If it is possible to have the data please send to my email address it is teddiyo.21@gmail.com
Hi Alan, thanks for this video! P.S. You taught me 10 years ago :)
c.educ##c.educ is like generating a quadratic effect? That is, its the same as gen edu2=educ*educ???
You did this because you assume that educ and prestige did not have a linear relationship but a quadratic one?
I can you all this margin commands after a logistic regression as well, right? It would be great if you do a video explaining margins (post estimations commands) for logistic regressions.
Thank you very much!
Marvin
+Marvin Aliaga Yes, that model was testing for a quadratic relationship. It is best to use factor notation so you can use margins and Stata can keep track of all your variables. This is particularly important with quadratics and logistic (or other limited dependent variables). When you use margins after a logistic regression, margins reports probabilities, which are easy to interpret. Then, you can use marginsplot to visualize the model. This is a real strength of Stata.
Best,
Alan
Hello, I have a question can I use the "margins at" command for values that are not in the sample? Let's say the max. education level in the sample is 10 years and I want to know the potential effect of 15 years, can I use "margins, at (15)" for that ?
Hi Tony, yes, this can be cautiously done, but be careful. Making predictive estimates outside the range of the data used in your model can produce nonsensical results. For example, based on General Social Survey data I can show you a model indicating an average decrease in sexual frequency of 1.2 times per year for each year of aging. This model probably works pretty well near the middle of the distribution of age, about 48 years old. This model includes respondents between the ages of 18 and 89 and may provide some information about octogenarians, but is not useful at all when examining people who are 100 years old!
Here is another example that demonstrates that you can use values outside of the range of your data in the margins command, but also produces nonsensical results. I used the auto dataset included with Stata to estimate a fuel efficiency model relating MPG to the weight of cars. Then, I used margins to estimate the average MPG for an M1 Abrams tank that weighs approximately 120,000 lbs., clearly outside the range of my data! Run this little program to see the results.
Best,
Alan
sysuse auto, clear
/* Estimate a simple model */
regress mpg c.weight##c.weight
/* Get some reasonable bounds for weight */
estat summarize
/* Margins within the range of the data */
margins, at(weight=(1800(200)5000))
marginsplot, recast(line) recastci(rarea) scheme(s1mono)
/* Margins outside the range using the weight of an M1 Abrams tank */
margins, at(weight=120000)
Thank you Alan for the quick and competent answer
@@bRabbit120793 👍
how do you do this for multiple independent variables?
Hi Jodi,
Yes, you can have two or more IV's in your model and can use them to estimate the marginal effects. It requires some thought about what you want to see from your model. I have provided some Stata code below that may give you some ideas about how to do this.
sysuse auto, clear
regress mpg weight displacement
estat summarize
margins, at(weight=(1800(200)4800))
marginsplot, recast(line) noci name(m1a, replace)
margins, at(displacement=(100(50)400))
marginsplot, recast(line) noci name(m1b, replace)
margins, at(weight=(1800(500)4500) displacement=(100(100)400))
marginsplot, recast(line) noci name(m2, replace)
regress mpg c.weight##c.displacement
margins, at(weight=(1800(500)4500) displacement=(100(100)400))
marginsplot, recast(line) noci name(m3, replace)
intro theme song was awesome
"The White Collar Holler" by the great Stan Rogers. Traditional music. My favorite song by Rogers is "Northwest Passage". I think Pierre Trudeau called it the unofficial national anthem of Canada!
Hi, I'm facing a problem on marginsplot ti my study. Is it possible to send you an e-mail with my problem? I thank you very much!!!
Sure, send me your problem and I can see if I have an answer or can send you in the right direction.
@@smilex3 Hi Alan, I have answered you two times, but I don't know why my message is being erased.