07-mle-and-mm

[sallymeeyan]

FAIL

• For MM of Weibull for gh, decrease your lower bound to 5 or even lower, because now it seems the result is right at the lower bound of 10, which indicates that it may not be 'optimal'.
• Your post needs better headings to help the audience to know which part is for which variable and method. For example, in 2 MLE parameters section, I assume the first chunk is for normal distribution using MLE, but it seems the second plot is for MM of Gamma distribution? Please use MLE for Gamma and Weibull distribution to re-do the part. You need a new nLL and fit for each of the two distributions. You can see from the plots the fit of Weibull is off - both pdf and CDF. The fit for Gamma is slightly better, but you did not use MLE and the estimates are also a bit off.
• In the first chunk of 2 MLE parameters section (for normal distribution), the qqplot is incorrect. You should use the same method you used for qqplot in the previous section. The following code has nothing to do with your sample, and the returned plot is just a perfect normal since you draw from a Normal(0,1) distribution. U <- runif(1000) Y <- qnorm(U) qqnorm(Y) abline(0,1)
• Use a new heading when you go to height, the new variable, and clearly label with the method you use - MLE or MM. Most of the parts of your work are hard to locate.
• For MM of height for normal, the pdf, CDF and QQplot are all off - I think you miss used variance s2 for sd in qnorm.
• For MM of height for Gamma, the pdf is not overlaid with histogram.
• For MLE of height for normal, the qqplot is off -see the above comment.
• What you have in the MLE section for height ( for Gamma and Weibull) is not MLE method. You need a new nLL and fit for each of the two distributions. Please redo the part. You can see from the plots the fit is very bad for the two.
• Overall, you need to better organize your post and add some comments, interpretation of the figures, comments of results, summary of findings, etc.
• Rename your work as writeup.html and writeup.rmd. The TAs use the same code to open every student's work so it's important that everyone uses the same file name.

[sallymeeyan]

FAIL

• No changes.
• If you have any questions about the feedback, please let me know. We can set up a meeting to walk through it.
• Though there seem a lot, many of your issues are easy to fix - I just listed everything I spotted but some of them are similar.

[sallymeeyan]

FAIL

• No changes.
• Just a reminder, the final due date is Wednesday, Dec 8th. The last chance to get feedback will be the regrading on Monday, Dec 6th.

[ZhuoyiZhan]

Hi, Yanyan:

I forgot to push to GitHub. Can you check my assignment please? I just resubmit.

Zhuoyi

[sallymeeyan]

FAIL

Thanks for the updates. Your report reads so much better than the first version! However, you're still having the following issues that I listed in the first grading:

• For MM of Weibull for gh, decrease your lower bound to 5 or even lower, because now it seems the result is right at the lower bound of 10, which indicates that it may not be 'optimal'.
• In the MLE for gh section (for all three distributions), the qqplots are incorrect. You should use the same method you used for qqplot in the MM section. The following code has nothing to do with your sample, and the returned plot is just a perfect normal since you draw from a Normal(0,1) distribution. U <- runif(1000); Y <- qnorm(U); qqnorm(Y); abline(0,1)
• Same thing for MLE of ht, all three qqplots are incorrect. It should be sample_quantile vs. theo_quantile, where sample quantiles are from ht, and theo_quantiles are from the distribution you estimated. Just re-use the code in your MM section. If you have difficulties understanding this, let me know so we can discuss about it.
• Rename your work as writeup.html and writeup.rmd. The TAs use the same code to open every student's work so it's important that everyone uses the same file name.

[ZhuoyiZhan]

Fixed it.

[sallymeeyan]

PASS

Very good.

• Just one minor issue - in MM-height-normal, you should report sqrt(s2) instead of s2, since sd is the parameter to be estimated.