11:55 The reason why your model is different every time is because you use the sample() function to get the random subset of markers. If you want a pseudorandom reproducible model, use set.seed(), for example set.seed(123) before the sample() function.
Hi Kevin. Very Nice video. We could really use more of these! I've just gotten to using rrBLUP and it can get pretty complicated. I've watched this video a few times and I think you're off on a couple of things you've said. It might also just be me, so I'm looking forward to hearing back from you. 1. You say that you are testing the accuracy of your model with cor() and say it's 61% accurate. But cor() is actually calculating linear correlation, which is not strictly speaking, 'accuracy'. It means your predicted values are in corcordance with the observed values and as a measure or index of the concordance, we can say the Pearson's correlation coefficient is 0.61. A p-value from cor.test() (which I can see you have commented out) would make easier interpretation of the significance of this correlation. 2. You said that you run your model 500 times because the BLUE and BLUP values are a little different everytime. But to my understanding, with the same markers (Z) and same phenotypes (y), mixed.solve() will always output the same results. If you are getting different BLUE and BLUP values for the same marker/phenotype combination (i.e. your defined training set), I suspect you are sampling multiple training sets. Which, to my understanding, is not good practice for testing your models.
This is really interesting, Sir. Please, how can I fit principal components in the mixed.solved model of rrBLUP to correct for possible population structure?
very informative! Can we if possible have access to the script. Also as suggested by @Ali Mohsin appreciate if you have a video for K-fold cross-validation
@@vladimirtorres1181 te deseo suerte. Pregunté hace 2 años jaja. No me respondió, pero no está tan difícil correr rrblup. No creo tener mis scripts a la mano, pero en estos días le echo un ojo y te digo.
11:55 The reason why your model is different every time is because you use the sample() function to get the random subset of markers. If you want a pseudorandom reproducible model, use set.seed(), for example set.seed(123) before the sample() function.
Thank you really helpful!!
Thank you for a nice rrblup introduction. Can you make one video more for K-fold cross-validation?
Hi Kevin. Very Nice video. We could really use more of these! I've just gotten to using rrBLUP and it can get pretty complicated.
I've watched this video a few times and I think you're off on a couple of things you've said. It might also just be me, so I'm looking forward to hearing back from you.
1. You say that you are testing the accuracy of your model with cor() and say it's 61% accurate. But cor() is actually calculating linear correlation, which is not strictly speaking, 'accuracy'. It means your predicted values are in corcordance with the observed values and as a measure or index of the concordance, we can say the Pearson's correlation coefficient is 0.61. A p-value from cor.test() (which I can see you have commented out) would make easier interpretation of the significance of this correlation.
2. You said that you run your model 500 times because the BLUE and BLUP values are a little different everytime. But to my understanding, with the same markers (Z) and same phenotypes (y), mixed.solve() will always output the same results. If you are getting different BLUE and BLUP values for the same marker/phenotype combination (i.e. your defined training set), I suspect you are sampling multiple training sets. Which, to my understanding, is not good practice for testing your models.
This is really interesting, Sir. Please, how can I fit principal components in the mixed.solved model of rrBLUP to correct for possible population structure?
very informative! Can we if possible have access to the script. Also as suggested by @Ali Mohsin appreciate if you have a video for K-fold cross-validation
Thanks for the useful video. Please also post R code on maybe Github. Thanks!
Thanks for your video, where can I get the r code?
Hi Kevin. Thank you very much!!. Could be possible to have acces to the scripts used in this tutorial?
Lol, I know that dude!!. Hola Eric!! iba a preguntar lo mismo, te respondió?
@@vladimirtorres1181 te deseo suerte. Pregunté hace 2 años jaja. No me respondió, pero no está tan difícil correr rrblup. No creo tener mis scripts a la mano, pero en estos días le echo un ojo y te digo.
Thank you Kevin very nice rrblup training. Would you please share me the script
I am very interesting to work on GS
Dear, I have enjoyed your talk but how could I get your script that you have been using for this tutorial?