Research and fun: Albert and Pera
Jonathan Helm
Two years we did all kinds of things that had nothing to do with Social Ccience Research. In 2017, the situation changed because there was still an unsatisfactory solution for the estimation of the model for experiments we had done to determine the quality of survey questions. Diana Zavalla of RECSM, who had learned at the UPC about Bayesian estimation procedures, had spoken with some experts in Bayesian methods about our estimation problems. This led to a joint paper: Helm, J. L., Castro-Schilo, L., Zavala-Rojas, D., DeCastellarnau, A. & Oravecz, Z. (2017). Bayesian Estimation of the True Score Multitrait–Multimethod Model with a Split-Ballot Design. Structural Equation Modeling: A Multidisciplinary Journal, 25:1, 71-85.
This paper suggested that the estimation problems of these models could be solved by use of Bayesian methods. To be honest, I thought that this development was going in the wrong direction. Bayesian approaches are useful for some problems but not for our problem. This was an identification problem of lack of information.
This paper suggested that the estimation problems of these models could be solved by use of Bayesian methods. To be honest, I thought that this development was going in the wrong direction. Bayesian approaches are useful for some problems but not for our problem. This was an identification problem of lack of information.
Albert Satorra
I needed the help of my friend Albert Satorra
Therefore, I went to my colleague and friend Albert Satorra to speak with him about this identification problem. In the past we had chosen for the solution to start with the assumption that the model across all countries would be the same. That was a very strong assumption but in this way we could start the estimation and slowly allow for differences in the models in the different countries. This approach required too many corrections and was therefore questionable.
Albert suggested a much less radical solution. In cross national research it is in general required that the respondents in all countries react approximately in the same way on our questions otherwise it is impossible to compare results of surveys across countries. This invariance requirement is a much less extreme restriction to start with than the one we had used before. In that case one can also expect that one can estimate the models in all countries and that fewer corrections in the model will be needed.
I tried this approach out on earlier collected data and it worked well. We published one paper and after that I analyzed the same data also with the Bayesean approach. In this way we could compare the results between the two methods. In a second paper, we could also show that our approach led to much smaller variation of the coefficients across countries than the Bayesian approach.
An invitation for a conference
Therefore, I went to my colleague and friend Albert Satorra to speak with him about this identification problem. In the past we had chosen for the solution to start with the assumption that the model across all countries would be the same. That was a very strong assumption but in this way we could start the estimation and slowly allow for differences in the models in the different countries. This approach required too many corrections and was therefore questionable.
Albert suggested a much less radical solution. In cross national research it is in general required that the respondents in all countries react approximately in the same way on our questions otherwise it is impossible to compare results of surveys across countries. This invariance requirement is a much less extreme restriction to start with than the one we had used before. In that case one can also expect that one can estimate the models in all countries and that fewer corrections in the model will be needed.
I tried this approach out on earlier collected data and it worked well. We published one paper and after that I analyzed the same data also with the Bayesean approach. In this way we could compare the results between the two methods. In a second paper, we could also show that our approach led to much smaller variation of the coefficients across countries than the Bayesian approach.
An invitation for a conference
This was not the end of the story because a bit later we received an invitation for a meeting between us and the Bayesian experts in a beautiful castle in France, at the coast near Nantes.
As the photo of the castle shows, it was really challenging to go there even if it was only for the nice surrounding. However because of the weak health of Irmtraud, we had to cancel already several times appointments, therefore we decided not to participate. Albert went to the meeting and Daniel Oberski substituted me in the meeting. They enjoyed it a lot and they told me that they had very interesting discussions.
As the photo of the castle shows, it was really challenging to go there even if it was only for the nice surrounding. However because of the weak health of Irmtraud, we had to cancel already several times appointments, therefore we decided not to participate. Albert went to the meeting and Daniel Oberski substituted me in the meeting. They enjoyed it a lot and they told me that they had very interesting discussions.
Then I met Yayo Pera
I was occasionally skating in an around Barcelona and at some point in time in 2017 I met Yoyo Pera as he calls himself (opa Pera). As you can see in the photo, it was difficult to miss him. He is two years older than I. He is also a fanatic skater, like me. He is not only a fanatic skater but also a convinced Catalan, very much in favor of independence. However, I think that he used this special outfit also to get in contact with people. On Sundays he goes to the large square next to the Arc de Triumph in Barcelona where many people are walking around. He is skating up and down the square and regularly people start to talk with him.
He is a nice person who has had a hard life in the past but now he enjoys the freedom he has when he goes skating. He told me that he also went regularly to the Barca complex where one can also skate in the ice hockey rink. I had not been there before but now we made an appointment to go to skate also on ice. There he was not wearing his spectacular dress but he, nevertheless, got a lot of attention of the children who skated there and of the personnel working there because he knows them already for a long time.
I was occasionally skating in an around Barcelona and at some point in time in 2017 I met Yoyo Pera as he calls himself (opa Pera). As you can see in the photo, it was difficult to miss him. He is two years older than I. He is also a fanatic skater, like me. He is not only a fanatic skater but also a convinced Catalan, very much in favor of independence. However, I think that he used this special outfit also to get in contact with people. On Sundays he goes to the large square next to the Arc de Triumph in Barcelona where many people are walking around. He is skating up and down the square and regularly people start to talk with him.
He is a nice person who has had a hard life in the past but now he enjoys the freedom he has when he goes skating. He told me that he also went regularly to the Barca complex where one can also skate in the ice hockey rink. I had not been there before but now we made an appointment to go to skate also on ice. There he was not wearing his spectacular dress but he, nevertheless, got a lot of attention of the children who skated there and of the personnel working there because he knows them already for a long time.
In the summer, we participated together in a championship in inline skating in Raval, a speed skating championship where I had been before with the speed skaters. Now I went there for fun with my friend Yayo Pera. Again we got of course a lot of attention being the oldest skaters who were participating. Parents wanted a foto of their children with these two old skaters. Later we also went together to skate along the Rio Besos , a tour of 10 km what was also a lot of fun seeing again some old speed skating friends. Pera was of course making a show again.
The difference between the first part and the second part of this story cannot be larger, but both activities I enjoyed and the people I had contact with were both very nice although very different. I have to add that I also have skated with Albert Satorra in Barcelona in the earlier years when we were here. But with him, after skating, we mostly spoke about statistical problems. That I did not do with Pera. But we had enough other topics to talk about.
