I have 20 independent variables regarding the practices and attitude of fishers about fisheries management. Data was collected though questionnaire and recorded as YES/NO. I would like to identify the variables that have less contribution towards fisheries management. Is there anyone who can suggest me what kind of analysis to use? Do you think that PCA is appropriate?
given the structure of your data (binary response) correspondence analysis may be more suited than PCA
packages in R : ca for a start, or the function corresp() in MASS
more with https://cran.r-project.org/web/views/Multivariate.html under the key "correspondence"
see also https://www.r-bloggers.com/correspondence-analysis-in-r/
best wishes
Given your data are binary, it might also be worth considering IRT. Like Jean says for correspondence analysis, there are lots of R packages for IRT. While correspondence analysis (the Gifi procedures) are PCA methods for ordinal and nominal data, IRT is a collection of procedures usually for binary and ordinal procedures from the same general model of factor analysis (see Bartholomew et al,'s textbooks on the general linear latent variable model).
Dears Jean Maccario and Daniel Wright
I thank you very much for your recommendations. I will go through all the recommended analysis and I will see the one, which fits to my data more.
Choosing between models based on which fits better is in general a risky procedure. There is a lot of backlash in medicine and psychology (and some oher disciplines, but because of their size these get picked on the most) about fishing for results that give you the results that you want. That said, the goals of PC and latent variable models are often both to simplify the patterns in the data to allow better understanding of the relationships. In this way, whichever presents the most simple story might be considered best. This was the advice my main grad school professor (David Bartholomew, so one of the more knowledgeable people in this solar system on the topic) gave.
\
Dear Daniel Wright
I thank you very much for your in-detail explanation. Of course I have no any information about latent variable models before, I will read it and use it for my data analysis. I thank you very much once again.
- Badges
- Science topic