I have three datasets given to me of images of a conveyor belt taken every 30 seconds over 3 days, but one image was removed/lost in one of them and a few (~10-20) in the others. I calculated the textural features of these images and my goal is to make a regression model to predict issues in conveyance (y). Can I do a normal time series analysis with this dataset? Or should I use the time in hours calculated and do a normal logistic regression with the time in hours being one of my covariates along with the textural feature and my Y(t-1) term? I am doing this analysis in R.
Usual Regression Analysis / Newton - Raphson forward / backward interpolation
formula .
Try TRAMO-SEATS program, dealing with outliers and missing data in time series analysis.
Usual Regression Analysis / Newton - Raphson forward / backward interpolation
formula .
You might also look at this book:
https://methods.sagepub.com/book/interrupted-time-series-analysis
Best, D. Booth
You can compute the autocorrelation (ACF) and partial autocorrelation (PACF) functions to check for time dependence and design your time-series model. Here you can find a matlab function for incomplete time-series:
https://doi.org/10.5281/zenodo.3518227
You can also simulate the missing data inside the time-series. Multiple-point statistics can handle well this kind of problems:
Article Missing data simulation inside flow rate time-series using m...
You can access the latest MPS simulation software here:
https://wp.unil.ch/gaia/downloads/
Best,
Fabio
Jaydip Datta
I have a question about your response -- do you mean by using usual regression analysis, that it is okay to ignore the missing data points and add time and a y(t-1) term in as covariates instead of storing data as a time series function? This is what I have done and I have gotten a good misclassification error when testing on another dataset. Is this acceptable to do?
- Badges
- Science topic
- Similar topics
- Computer Science
- Images