STAR WARS : Episode IX December 20, 2019 6 6

[valiantman]

19 minutes ago, Bosco685 said:

42 minutes ago, valiantman said:

I could point out your parentage includes a hamster and the scent of elderberries, but that's a much better comedy that your comedy of errors.

'than' 

Now, we have a fantastic learning opportunity.  You have pointed out my obvious mistake. I now thank you and go back and correct my mistake. 

Thank you.

I have corrected my obvious spelling error.

 

 

 

 

 

 

[Bosco685]

46 minutes ago, valiantman said:

Now, we have a fantastic learning opportunity.  You have pointed out my obvious mistake. I now thank you and go back and correct my mistake. 

Thank you.

I have corrected my obvious spelling error.

 

Good on you!

Just like when I threw together a quick chart of the IMDb data and upon further modification from a binary to a three-score system as part of our discussion you noted one chart resulted in the wrong percentage. I took it in stride, and joked Star Wars fans give 114%. I didn't take offense. But you implying my 'panties twisted' over things like this. Meanwhile, there was no misrepresentation of the audience score nor the critic score.

You ASSUMED it was meant to distort details. Yet the critic score with a range of 52% to 58% has an axis of 50% to 60% because the range is so tight. So now even with a 80% to 90% axis on the audience score, the result is still the same. What a win!

The propaganda campaign revealed!

[valiantman]

1 hour ago, Bosco685 said:

 

Stop changing the first chart. Your error is in the second chart. You need to show the drop from 58% to 52% on a scale which has the same minimum and maximum values as the first chart.  0% to 100% is the most appropriate for both charts.  The "drastic fall" you're trying to show from 58% to 52% will be appropriate at 6%, not a fall from high on the chart to low on the chart. The other four errors will remain, but you will have fixed the worst one.

The fact that you chopped the first chart down to a different 10-point range in a different part of the 0% to 100% did not improve either chart.

Edited February 1, 2020 by valiantman

[valiantman]

4 hours ago, Bosco685 said:

Oh, and guess what? Consistent 86% even now.

Any sample size of 8,158 (the very first data point on the chart) with a 86% average has a statistical 99.99999% confidence that the actual value for the full population is between 85.8% and 86.2%.  

As your black belt tells you, Six Sigma is a confidence of 99.99966%, so we're already better than Six Sigma from the very first data point that the answer is going to remain between 85.8% and 86.2% forever.

EVERY larger data point after the first data point will have even more confidence that the correct number is between 85.8% and 86.2%.

AT WHAT POINT, specifically, would you expect the value to fluctuate to something OTHER than 86%? When should it mathematically be even 85% or 87%?

You're a fellow statistician... enlighten us.

Edited February 1, 2020 by valiantman

[Bosco685]

1 hour ago, valiantman said:

Stop changing the first chart. Your error is in the second chart. You need to show the drop from 58% to 52% on a scale which has the same minimum and maximum values as the first chart.  0% to 100% is the most appropriate for both charts.  The "drastic fall" you're trying to show from 58% to 52% will be appropriate at 6%, not a fall from high on the chart to low on the chart. The other four errors will remain, but you will have fixed the worst one.

The fact that you chopped the first chart down to a different 10-point range in a different part of the 0% to 100% did not improve either chart.

That's okay. They stay as-is. Kind of like a fixed Rotten Tomatoes audience score. 

[Broke as a Joke]

This thread was more fun when we all agreed about how bad a movie this was.  

[Bosco685]

Just now, valiantman said:

Any sample size of 8,158 (the very first data point on the chart) with a 86% mean has a statistical 99.99999% confidence that the actual value for the full population is between 85.8% and 86.2%.  

As your black belt tells you, Six Sigma is a confidence of 99.99966, so we're already better than Six Sigma that the answer is going to remain between 85.8% and 86.2% forever when we have our first data point at 8,158.

EVERY larger data point after the first data point will have even more confidence that the correct number is between 85.8% and 86.2%.

AT WHAT POINT, specifically, would you expect the value to be something OTHER than 86%.  You're a fellow statistician... enlighten us.

Actually, you just confirmed with that statement you are not clear on the practice of Six Sigma process design. Although in an ideal environment you would want to achieve the goal of 3.14 defects per a million opportunities, the confidence interval set as a goal will be based on the stakeholder's risk appetite and the organization constraints (time, resources, quality standards). What you did is applied best-practice statistical confidence interval (99%, 95%), but when applying Six Sigma as an business operations practice you can't just force working past the business constraints. You can present what could be operational goals. But if the Voice of Customer (requirements) and the Voice of Process (process capabilities) can only achieve statistical alignment to a certain confidence interval, then anything past this is noted as the process variance.

Applies this to a survey result as if that is the Six Sigma standard was an error as we would need to know what the VOC is for satisfactory statistical results. But hey - you attempted to get another troll barb in there. And then assumed adding 'Six Sigma' made it more relevant.

[paperheart]

8 minutes ago, Broke as a Joke said:

This thread was more fun when we all agreed about how bad a movie this was.  

this thread is now more entertaining than the movie ever was

[valiantman]

11 minutes ago, Bosco685 said:

36 minutes ago, valiantman said:

Any sample size of 8,158 (the very first data point on the chart) with a 86% mean has a statistical 99.99999% confidence that the actual value for the full population is between 85.8% and 86.2%.  

As your black belt tells you, Six Sigma is a confidence of 99.99966, so we're already better than Six Sigma that the answer is going to remain between 85.8% and 86.2% forever when we have our first data point at 8,158.

EVERY larger data point after the first data point will have even more confidence that the correct number is between 85.8% and 86.2%.

AT WHAT POINT, specifically, would you expect the value to be something OTHER than 86%.  You're a fellow statistician... enlighten us.

Actually, you just confirmed with that statement you are not clear on the practice of Six Sigma process design. Although in an ideal environment you would want to achieve the goal of 3.14 defects per a million opportunities, the confidence interval set as a goal will be based on the stakeholder's risk appetite and the organization constraints (time, resources, quality standards). What you did is applied best-practice statistical confidence interval (99%, 95%), but when applying Six Sigma as an business operations practice you can't just force working past the business constraints. You can present what could be operational goals. But if the Voice of Customer (requirements) and the Voice of Process (process capabilities) can only achieve statistical alignment to a certain confidence interval, then anything past this is noted as the process variance.

Applies this to a survey result as if that is the Six Sigma standard was an error as we would need to know what the VOC is for satisfactory statistical results. But hey - you attempted to get another troll barb in there. And then assumed adding 'Six Sigma' made it more relevant.

I wanted you to feel like I listened to you when you threw around your credentials since you have thrown mine around a dozen times since I did once.  I care nothing for "Six Sigma" as a vendor, but I care about six sigma as a confidence interval.  As you mentioned, normal confidence intervals are 99%, 95%, but they can be calculated for any number.  Therefore:  Any sample size of 8,158 (the very first data point on the chart) with a 86% mean has a statistical 99.99999% confidence that the actual value for the full population is between 85.8% and 86.2%.  

...and we're back to...

AT WHAT POINT, specifically, would you expect the value to be something OTHER than 86%.  You're a fellow statistician... enlighten us.

[Bosco685]

1 hour ago, valiantman said:

I wanted you to feel like I listened to you when you threw around your credentials since you have thrown mine around a dozen times since I did once.  I care nothing for "Six Sigma" as a vendor, but I care about six sigma as a confidence interval.  As you mentioned, normal confidence intervals are 99%, 95%, but they can be calculated for any number.  Therefore:  Any sample size of 8,158 (the very first data point on the chart) with a 86% mean has a statistical 99.99999% confidence that the actual value for the full population is between 85.8% and 86.2%.  

...and we're back to...

AT WHAT POINT, specifically, would you expect the value to be something OTHER than 86%.  You're a fellow statistician... enlighten us.

Six Sigma as a confidence interval (3.4 DPMO = 99.99966%) is a goal. But a process doesn't always achieve that goal due to business constraints. Asking me when a survey will achieve Six Sigma isn't the question to be asked. That's not how the practice works when you are measuring defects per million opportunities. Then stating you just threw it out there to play at acknowledging my credentials only further confirms the intent of the question.

Now, I think how you are attempting to statistically address this situation assuming it is a sample survey of a set population. That's not how Rotten Tomatoes and IMDb works. They don't determine the population potential to see a film, and then measure based on a 95% or 99% confidence intervals. Otherwise, they would have to apply data from sources like the World Bank and based on the estimate the average audience member will be 15-64 (65.33% of the overall world population) as a target potential. Then based on the population (Population = 4,961,160,200) and RT sample size (96,493) you would be able to determine your margin of error of this sample being representative of the overall population. In this case, the margin of error would be...

It's a game you are playing to diminish my stats abilities, and reestablish your own. Meanwhile, to state an 86% constant positive across 96,493 contributors is statistically sound is interesting. Tell me, fellow statistician, knowing this is satisfaction survey where each user has to vote 3.5/5 for the result to be POSITIVE and anything less is ROTTEN at what point would it be 86%? The answer right now based on the data we are presented is all day long, no matter the count.

Edited February 1, 2020 by Bosco685

[Bosco685]

Satisfaction survey results to date:

[valiantman]

2 hours ago, Bosco685 said:

Six Sigma as a confidence interval (3.4 DPMO = 99.99966%) is a goal. But a process doesn't always achieve that goal due to business constraints. Asking me when a survey will achieve Six Sigma isn't the question to be asked. That's not how the practice works when you are measuring defects per million opportunities. Then stating you just threw it out there to play at acknowledging my credentials only further confirms the intent of the question.

Now, I think how you are attempting to statistically address this situation assuming it is a sample survey of a set population. That's not how Rotten Tomatoes and IMDb works. They don't determine the population potential to see a film, and then measure based on a 95% or 99% confidence intervals. Otherwise, they would have to apply data from sources like the World Bank and based on the estimate the average audience member will be 15-64 (65.33% of the overall world population) as a target potential. Then based on the population (Population = 4,961,160,200) and RT sample size (96,493) you would be able to determine your margin of error of this sample being representative of the overall population. In this case, the margin of error would be...

You were doing so well!  Then you stopped typing and just put in ...

 

The answer to your question is:

 

The Margin of Error (MOE) is calculated according to the formula: MOE = z * √p * (1 - p) / √(N - 1) * n / (N - n)

Where: z = 2.576 for a confidence level (α) of 99%, p = proportion (expressed as a decimal), N = population size, n = sample size.

z = 2.576, p = 0.86, N = 4961160200, n = 96493

MOE = 2.576 * √0.86 * (1 - 0.86) / √(4961160200 - 1) * 96493 / (4961160200 - 96493)

MOE = 0.894 / 310.636 * 100 = 0.288%

The margin of error (with finite population correction) is ±0.288%

 

That means the answer is 86% with all four billion people (you picked the number this time) being between 85.7% and 86.3%.

It's 86% until every person on the planet in the population you selected has voted.

 

You're right, that was fun.

[Bosco685]

1 hour ago, valiantman said:

You were doing so well!  Then you stopped typing and just put in ...

 

The answer to your question is:

 

The Margin of Error (MOE) is calculated according to the formula: MOE = z * √p * (1 - p) / √(N - 1) * n / (N - n)

Where: z = 2.576 for a confidence level (α) of 99%, p = proportion (expressed as a decimal), N = population size, n = sample size.

z = 2.576, p = 0.86, N = 4961160200, n = 96493

MOE = 2.576 * √0.86 * (1 - 0.86) / √(4961160200 - 1) * 96493 / (4961160200 - 96493)

MOE = 0.894 / 310.636 * 100 = 0.288%

The margin of error (with finite population correction) is ±0.288%

 

That means the answer is 86% with all four billion people (you picked the number this time) being between 85.7% and 86.3%.

It's 86% until every person on the planet in the population you selected has voted.

 

You're right, that was fun.

Ohhhh. How impressive. And I left it open for your answer. Though again with incomplete data based on what Rotten Tomatoes provides, validating the accuracy is the big gap in any analysis. We have to trust in 86% users rated the film 3.5/5 or better.

And yet the answer is always 86%. Because that is the results we are presented. Guess data quality validation doesn't count. Go figure!

[mattn792]

http://en.wikipedia.org/modssaveusplease

[valiantman]

30 minutes ago, Bosco685 said:

And yet the answer is always 86%. Because that is the results we are presented. Guess data quality validation doesn't count. Go figure!

How do you suggest we check the data quality when every known method shows it is within a statistically-sound possibility?

You're not just presenting data, you're drawing a conclusion that includes accusing the studios of manipulation.  You can't justify your accusations, except to say "that's quite a coincidence" when the math says it's not.

Since it isn't propaganda, and I will allow for it to be something other than propaganda, but it also isn't coincidence, the math says so. So, your accusations are based on... gut feeling?

Let's data quality check those intestines, I guess.

[valiantman]

34 minutes ago, mattn792 said:

http://en.wikipedia.org/modssaveusplease

Despite the obvious animosity, @Bosco685 and I are a lot closer to middle ground than it seems.  Since this has been a discussion after the 108th page of the topic, I feel no guilt having monopolized the conversation which was otherwise dwindling to the bottom of the board.  Since we're entering the American political season, there is much overlap between audience survey and voter polls. The average journalist has about 5% of the statistical skill of @Bosco685, yet they will dominate headlines with their analysis of about a dozen variables they will roll up into a single number... poorly.  None of that political discussion belongs on the CGC Board, but the critical thinking applies universally.

 

Yes, I did compliment @Bosco685 multiple times in this post.  It's how I debate and I can take the jokes and animosity aimed at me, because a real internet argument without humor is not fun for anyone.

This was at least fun for me, so I got what I wanted.  Maybe others have also enjoyed an otherwise dead topic for a day or so. 

 

The horse is dead whether it gets beaten or not, so you might as well beat a dead horse with a rubber chicken.

 

Edited February 2, 2020 by valiantman

[Bosco685]

15 minutes ago, valiantman said:

How do you suggest we check the data quality when every known method shows it is within a statistically-sound possibility?

You're not just presenting data, you're drawing a conclusion that includes accusing the studios of manipulation.  You can't justify your accusations, except to say "that's quite a coincidence" when the math says it's not.

Since it isn't propaganda, and I will allow for it to be something other than propaganda, but it also isn't coincidence, the math says so. So, your accusations are based on... gut feeling?

Let's data quality check those intestines, I guess.

Actually I did justify my conclusion based on similar events where studios to counter a Rotten critic result potentially counter-marketed the result by tampering with the audience score. And in the case of the Gotti film it was so extreme, analysts caught it rapidly. So there are similar cases. You just glossed over this.

So in this case, not only did ROS land the lowest word-of-mouth score (B+), one of the Iowest Metacritics score (53/100), but also the lowest Rotten Tomatoes critic score (52%). All key data points clearly indicate the film was receiving mixed results. Forcing studio marketing action to avoid such data to disrupt audience attendance.

So much data that has nothing to do with propaganda. Or I guess you assumed none of that factors into the equation.

[mattn792]

If only PMs existed.

[valiantman]

2 minutes ago, Bosco685 said:

Actually I did justify my conclusion based on similar events where studios to counter a Rotten critic result potentially counter-marketed the result by tampering with the audience score. And in the case of the Gotti film it was so extreme, analysts caught it rapidly. So there are similar cases. You just glossed over this.

So in this case, not only did ROS land the lowest word-of-mouth score (B+), one of the Iowest Metacritics score (53/100), but also the lowest Rotten Tomatoes critic score (52%). All key data points clearly indicate the film was receiving mixed results. Forcing studio marketing action to avoid such data to disrupt audience attendance.

So much data that has nothing to do with propaganda. Or I guess you assumed none of that factors into the equation.

These are simply different measures, and vastly different individuals.  Professional critics should always have their motives questioned when just 10 people can cause significant swings in an overall score.  There are no 1,000 voters on the audience side which could have anything like the influence that 10 critics have.

 

Alternative hypothesis: Critics which dislike a movie universally liked by audiences and other critics are outliers, possibly because they genuinely like being in the spotlight even if they're wrong.

However, when critics approach 50%/50%, there is no way to ensure a larger portion of the spotlight, and critics fall back on the "image" they hope to convey to their public, "hipster", "fanboy", "fangirl", "curmudgeon", etc., and it is entirely possible that the total population of critics allowed to be part of the official critic score looks nothing at all like the audiences for these movies.  It is also likely that the critics two years ago aren't the same as the critics five years ago or today.  That's literally fishing in three different ponds and assuming the nets will pull in the same mix of fish. Unlikely.

 

Too long; didn't read:

If we do need to be suspicious of 500 of one thing or 90,000 of another, it's the 500 where each individual fish has a stinkier portion of what we're being served.

[valiantman]

1 minute ago, mattn792 said:

If only PMs existed.

I once watched a debate in a big room for debates with a sign over the door that said "Debates Here Every Day For 18 Years In A Row" and I thought to myself, "I should tell those debaters and all the other people in the room for debates that they should go debate somewhere else without any people."

 

No, wait.  I didn't think that.  That would have made me look ridiculous to everyone.