Hi Gisela,

SEM - PLS (Partial Least Square - with SMARTPLS, for example) allow you to use small samples when the causal model have only one predictor. With one predictor, under the conditions suggest by Cohen [Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences. 2nd ed. New York: Psychology Press.] you will need 55 respondents to calculate the SEM-PLS.

For more details see: FAUL, F.; ERDFELDER, E.; LANG, A.-G.; BUCHNER, A. G*Power 3: A flexible statistical power analysis for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, p. 175-191, 2007.

If you want donwload the freewear software for sampling calculatiuon:

http://www.gpower.hhu.de/en.html

Send us more information about your research. Do you will make comparisions with 3 or 4 respondents?

hope this help

best regards

Dirceu

What exactly are you looking for? Best, DEB

Dear Gisela,

The required sample size is related to the complexity of your model. Instead of using Classical theory, you could try Bayesian approach, which requires smaller sample size. You may have a look at Mplus manual guide in which you have some information of how to conduct Bayesian analysis.

On top of that, you can look at the work of Rens van de Schoot (https://scholar.google.com/citations?user=zclEqP8AAAAJ&hl=en) and Joop Hox (https://scholar.google.com/citations?user=M1pZvM4AAAAJ&hl=en) for more information regarding Bayesian analysis.

It would also be helpful if you give us more information, regarding the type of model, variables, scales so far and so on. It is quite difficult to give advice in the correct statistic analyses without having information, such as complexity of the model, sample size and topic of research. All those variables will play a role in deciding which model to use. In addition, we should make clear your research question, since it will impact the methodology that you have to use.

Best

dario

Hi Gisela,

SEM - PLS (Partial Least Square - with SMARTPLS, for example) allow you to use small samples when the causal model have only one predictor. With one predictor, under the conditions suggest by Cohen [Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences. 2nd ed. New York: Psychology Press.] you will need 55 respondents to calculate the SEM-PLS.

For more details see: FAUL, F.; ERDFELDER, E.; LANG, A.-G.; BUCHNER, A. G*Power 3: A flexible statistical power analysis for the social, behavioral, and biomedical sciences. Behavior Research Methods, 39, p. 175-191, 2007.

If you want donwload the freewear software for sampling calculatiuon:

http://www.gpower.hhu.de/en.html

Send us more information about your research. Do you will make comparisions with 3 or 4 respondents?

hope this help

best regards

Dirceu

Thank you very much for your responses. To be honest, I was a bit scared to expose my situation considering I am in a rush with my thesis project now I'm about to 8 months approximately to finish my master. Of course I can give more information of my research.

My original idea was to send questionnaires to mexican SMEs which have been involved - in some way - to the Green supply chain in order to prove that my variables (in these case, green practices) are related and therefore necessary to apply if companies would like to improve their economic, operational performance and being responsible with the environment as well. In my country there's a significant opportunity regarding on green topics, contrasting to the EU being leader on environmental subjects.

Then I realized it was going to be if not impossible, challenging to obtain that specific information that would be helpful for my investigation and useful for SEM. I may sounded a bit pesimistic but I felt that way at the beginning. With apparently no thesis director support, I started thinking on changing the approach of the investigation and using AHP just to priorize the most important variables and that was it, in a comparative study of 3 or 4 observations with all the requirements I am looking for. Today I had a meeting with my thesis director and naturally I was told I cannot use tools such as AHP because it is one of the most used at our master, and I must accept the situation.

So, my conclussion at this time is: I will start sending my survey to every single SME I might find, not even in my state but in the country, and see if I can reach the minimum sample necessary in order to proceed with SEM. I know I cannot use inferential statistics if I use nonrandom observations, that is why I'll do it randomnly. One good thing is, if even most of the companies respond negatively and confirm that they do not implement green actions in their companies, is also valuable information for my research.

I apologize for the extensive story, I needed to relieve myself.

Thank you and have a nice day.

Gisela,

Comments made by prof. Dirceu da Silva are quite useful and I already used SmartPLS for samples of 30 and 45 respondents with pretty good results.

Attached are some articles about SmartPLS, including two done by prof. Dirceu, which can be useful.

Regarding sample collection, for my thesis I used the following strategy: I went to LinkedIn and look for those companies which should fit in my research. After that I look for some people inisde the company who could help me on the research and send them an invitation for network, quickly explaining what was the purpose of inviation. As soon as they accepted me I returned with a more detailed message talking about the research.

The beauty of this approach was that as far as I have people in my net I could make some direct follow-up with them (worst case scenario was 3 f-ups) and "make" them answer the search.

Despite the fact my research was related to Sales area and guys in this field were my target, differently from yours, I got 268 answers which fit in my needs at that time (for my thesis I used AMOS instead of SmartPLS).

See my story as a suggestion of strategy to get answers.

Regards,

Alcides Barrichello

Thank you so much. I am checking right now the material you attached.

I really appreciate it.

Gisela

Hi Gisela,

When we use a survey with "politic correct" problematic. how is your case (environment. green causes, sustainability etc.). We can have a situation that implies em social desirability (the respondet show that we expect or want).

To minimize this situation you can ask in two ways:

1. Ask the respondent to show their attitude toward the item of the scale.

2. Ask the respondent to show how other people or professionals in the sector / segment would respond, that is, make the choice as if it were the others.

Answer 2 is the one he believes the most! Try it! Can will be more interesting.

Best regards

Dirceu

Dear Professor Da Silva!

First of all, I really appreciate the time you invested on advising me. Also I would like to express the hapinness I experienced once I read your paper about the Structural Equation Modelling, is a very complete article that can be easily understandable for beginners like me.

About your suggestion, is an interesting point what you showed about how could be biased the survey. I am actually running a pilot test in order to discover what other kind of ammendements I can do to the questionnaire. Even though I lauched it yesterday, I am on time to structure the questionnaire in a way the bias could be dimished in some way.

If you have some time and you would like to make your observations, I can attach the survey. Is in Spanish, so I hope it will not be a trouble.

https://www.questionpro.com/t/AN5EXZbRXk

Gracias :)

Warm Regards,

Gisela

Leí y hablo bien en español. No hay problema. Voy a leer con todo el cuidado. saludos

Me imaginaba que hablaba español también :), pero me daba pena preguntarle. Qué bien. Muchas gracias Profesor. Que pase muy buen día!

Gisela

Ola Gisela,

He leído con atención los elementos del cuestionario y tengo dos sugerencias de cambio.

Quédate completamente a gusto para aceptarlas o no.

12. Se controlan las emisiones de agentes contaminantes (tales como el dióxido de carbono CO2) que emite el transporte de la empresa

sugerencia de cambio:

12. Se controlan las emisiones de gases de efecto invernadero (tales como el dióxido de carbono CO2) que emite el transporte de la empresa

17. Reducción de emisiones de agentes contaminantes (tales como el dióxido de carbono CO2)

sugerencia de cambio:

17. Reducción de emisiones de gases de efecto invernadero (tales como el dióxido de carbono CO2).

Saludos

Dirceu

Oi Prof Dirceu!

Eu vou tentar falar na sua lingua :), muito obrigada por suas recomendações! Eu já fiz as modificações!.

Muchas gracias :) Pase un bello día

Saludos Cordiales

Gisela,