October Forum Survey Results

[GarrettBlake]

I would like to get your feedback on this thread
https://www.f150lightningforum.com/...charge-speed-on-dash.31565/page-4#post-580997

This is an example of a feature that was added in 24 I believe but has not made it via OTA for 22 and 23. It seems pretty straightforward and certainly useful. Is this anything we can expect in an OTA or will people have to go to the great lengths described in the thread to do it themselves?

Sponsored

 

[flyct]

A rough approximation of Forum members = about 21,000 from what I can see.
but the survey had only 200 replies. one member in 100. That seems surprisingly poor to me.
Perhaps the Admin know how many members are actually active on the forum, and the survey response of active members is a far better percentage than total membership implies.

A sample size of 100 from a population of 21,000 is actually statistically significant. A 100 reply sample size from a 21k population will probably net a ±10% Margin of Error.

That is from my Six Sigma Certification days.

 

[Jim Lewis]

That is from my Six Sigma Certification days.

You might want to check your notes from your Six Sigma classes... The following is from the Wikipedia article on Margin of Error, where in the CONCEPT section, n is defined as the sample size and N as the population size. The article notes that for large sample sizes (usually more than 50 to 100 samples, IIRC), the error distribution approaches a normal distribution. You can see that the margin of error basically increases with variability within the population (sigma squared) and decreases with increasing sample size. N, the population size, does not appear in the error calculation. The z factor relates to the degree of confidence one wants in the margin of error result (MOE).
Margin of error - Wikipedia section on CONCEPT.

The only time N gets involved, for the most part, is when the sample size approaches 5% to 10% of the population size, because then, as you go along, you start to change the population composition remaining to be sampled by the samples you have already gotten. That's explained in the same article in the Margin of error - Wikipedia section on the Finite Population Correction (FPC).

One can easily calculate that, for a binary survey question and a sample size of 100, the maximum % error will be about 10% without invoking the population size (as in your example). The maximum variance in a binary choice is when the responses are split ~50-50. Sigma squared is then 0.5 * 0.5 = 0.25. With n = 100, 0.25/100 = 0.0025, and its square root is 0.05. For (two-sided) 95% confidence, the z factor is ~2, so 2 * 0.05 = 0.10, or a 10% margin of error. The calculation doesn't involve the population size at all. A sample size of 100 divided by a 21,000 population size is ~0.5%, well below the sample size proportion (5% to 10%) where the Finite Population Correction needs to be invoked.

One can also ask ChatGPT or Google Gemini and get the same explanation (with less math!) that's contained in the Wikipedia Margin of Error article.

 

[flyct]

A sample size of 100 from a population of 21,000 is actually statistically significant. A 100 reply sample size from a 21k population will probably net a ±10% Margin of Error.

That is from my Six Sigma Certification days.

You might want to check your notes from your Six Sigma classes... The following is from the Wikipedia article on Margin of Error, where in the CONCEPT section, n is defined as the sample size and N as the population size. The article notes that for large sample sizes (usually more than 50 to 100 samples, IIRC), the error distribution approaches a normal distribution. You can see that the margin of error basically increases with variability within the population (sigma squared) and decreases with increasing sample size. N, the population size, does not appear in the error calculation. The z factor relates to the degree of confidence one wants in the margin of error result (MOE).
Margin of error - Wikipedia section on CONCEPT.

The only time N gets involved, for the most part, is when the sample size approaches 5% to 10% of the population size, because then, as you go along, you start to change the population composition remaining to be sampled by the samples you have already gotten. That's explained in the same article in the Margin of error - Wikipedia section on the Finite Population Correction (FPC).

One can easily calculate that, for a binary survey question and a sample size of 100, the maximum % error will be about 10% without invoking the population size (as in your example). The maximum variance in a binary choice is when the responses are split ~50-50. Sigma squared is then 0.5 * 0.5 = 0.25. With n = 100, 0.25/100 = 0.0025, and its square root is 0.05. For (two-sided) 95% confidence, the z factor is ~2, so 2 * 0.05 = 0.10, or a 10% margin of error. The calculation doesn't involve the population size at all. A sample size of 100 divided by a 21,000 population size is ~0.5%, well below the sample size proportion (5% to 10%) where the Finite Population Correction needs to be invoked.

One can also ask ChatGPT or Google Gemini and get the same explanation (with less math!) that's contained in the Wikipedia Margin of Error article.

Jim,

Thank you for confirming my posting that if n=1 the MOA is about 10%. Most "normal" people wouldn't understand the formula, like the two of us nerds do.

Sort of like trying to explain what the 68-95-99.7 Rule "One, Two and Three standard deviations (σ) of the mean" is to a normal person. They don't care! They just know that they love their Lightnings, as do we nerds.

Sponsored