Hi! Thanks for checking out my video on simple linear regression in SAS. Let me know what you thought and if you found it helpful. Also, let me know if you have any questions. Have an awesome day! Andy
Excellent! Glad to hear this tutorial was helpful to you! You can also find other helpful support resources on SAS Viya on our support pages 2.sas.com/6054H8pnM
I like the idea (I did not watch the entire video because I feel I already know linear regression) and I especially like that it was posted over at SAS Communities (but I'm not allowed to comment on the post there, so I comment here). SAS Communities needs more basic statistics and basic programming tutorials; there are plenty of tutorials that appear at SAS Communities on advanced topics like Hierarchical Models and Quadratic Discriminant Analysis and so on; this is the first I have seen at SAS Communities on basic statistics. Furthermore, there needs to be some mechanism to search for tutorials at SAS Communities (if there is such a mechanism, I am not aware of it) and there also needs to be some mechanism to search for VIDEO Tutorials at SAS Communities. And maybe even post a comprehensive list of all such video tutorials at SAS Communities (and keep re-posting the list every week, because there are always new people joining, or searching for something).
Thanks for the feedback Paige! We have a big list of topics to cover in this format. If you have ideas for the top statistics topics to tackle first, let us know!
Hi there Paige, we hear you on adding a mechanism to find video tutorials on the SAS Communities and are exploring ways to do that. Thanks for your input!
@@chrishemedinger1382 can you please do a multiple linear regression including interaction/ effect modification, then latent class analysis, trajectory analysis. Thanks!
"Simple"....Typical with SAS. NOTHING is simple. Everything is super complicated. More info than you need and all sorts of exceptions and that you just have to "know".
If there is a specific question you have, please let us know. If you have specific product feedback to help us in future product development, please feel free to share it on the SAS Idea Exchange: 2.sas.com/6050JQIRE. Additional info on customer feedback, here: 2.sas.com/6051JQIR1
I think there is a mis-statement at about 13:20 of the video: "P-value very small tells me that this model is doing a very good job at explaining a lot of the variability of the target." But that's not what the p-value tells you. It tells you that the fitted line is statistically significant, or in layman's terms that the fitted line did not happen by random chance. The r-squared tells you when the model is fitting a lot of the variability. You could have a p-value of 0.001 and a low r-squared, indicating that the fitted line is statistically significant but it does not explain a lot of the variability of the target. These concepts should be intertwined as the speaker said.
Hi Paige, I wouldn’t disagree with you. I made quite a few leaps for that general statement, but I was trying to avoid a deep down discussion for this video. The p-value for that ANOVA table in a linear regression is a test of the null hypothesis (that all the regression coefficients are 0; in this case we only have one.) A very small p indicates that we found evidence that at least one of our coefficients was non-zero. You can then look at the table of the parameter estimates to see that the p-values for both the y-intercept and our one term were also very small (also indicating evidence of being non-zero.) Thus, we found some inputs that appear to be statistically significant. I probably meant to throw in a small discussion about the r-square value, which is a much better statistic and indicating that the model is explaining the variability in the target. Thanks for catching that!
Hi John, thanks for your question. At time stamp 28:51, take a look to the right of the R-Square. You'll notice that out of approximately 1 million observations, only about 20% (211,509) are used in this linear regression. In this VS_Bank table, non-purchases are coded as missing for the target tgt Interval New Sales. So basically, this linear regression is only using 20% of the original data. I think this is the best explanation for why that R-Square is so low. I hope this helps! Thanks, Andy
@@andyravenna4222 thanks for replying. I am a new student to linear regression and SAS. So based on your explanation, if the data is cleaned by removing empty sales, the RSquare will increased?
@@heahkl Hi John, it is great that you are so interested in Linear Regression. Since you are new, I'd like to encourage you to take our Statistics 1 course: support.sas.com/edu/schedules.html?crs=STAT1&ctry=US . It is an Introduction to ANOVA, Regression, and Logistic Regress. To get back to your new question, it turns out that this particular data table has already been cleaned. Because those sales values are missing, they are ALREADY being removed from the Linear Regression. If you are looking for ways to increase the R-Square, we would have to consider other alternatives, such as adding interaction terms or considering an alternative model. The Statistics 1 covers a lot this. I hope this helps! Thanks, Andy
