You can also use statistics for actual factual arguments and deductive reasoning. If a statistic that is well founded says that >60% of people questioned think X about Y, then that is a good foundation to make the argument that "yes, a majority of people seem to think X about Y." So... Don't assume that everyone who cites a statistic does it just to lie. The absence of any facts whatsoever makes for debating, and we could all use a little less of that these days.
RPG Maker Forums
Galenmereth
Tai_MT
The problem with that argument, Galen, is that you cannot reasonably assume anything about your statistics based on anything less than the full human populace. There are any number of ways to manipulate your statistics, especially by picking and choosing who you put in your study (or "what"). Anytime someone cites statistics they should issue the disclaimer saying that their stats may not be accurate at all.
Ms Littlefish
Yes, I agree.
Numbers are not always used to lie and the ability to make models and gather data is pretty much integral to every field of science.
But I think it's healthy to be skeptical of big claims (especially when it comes to media coverage) and admit that tests can be poorly designed.
Numbers are not always used to lie and the ability to make models and gather data is pretty much integral to every field of science.
But I think it's healthy to be skeptical of big claims (especially when it comes to media coverage) and admit that tests can be poorly designed.
Tai_MT
Case in point: "Statistics and studies show that 100% of people who eat vegetables die." Of course that's factual, but it can be very misleading. See my point about statistics being the most dangerous form of lie ever created?
mlogan
I find that 83% of people who use statistics are making them up.
Ms Littlefish
@ Tai_MT the crucial aspect of any successful statistical analysis is building a good sample size and designing the right test. Get that part wrong and the test will likely produce inaccurate results.
It's not possible to test the entire population and that's why there is a lot of care in setting up an experiment.
The other half of the battle is knowing how to interpret your data.
If someone has manipulated their test, it's a bad test. Some grou...
It's not possible to test the entire population and that's why there is a lot of care in setting up an experiment.
The other half of the battle is knowing how to interpret your data.
If someone has manipulated their test, it's a bad test. Some grou...
Ms Littlefish
Some groups will use bad tests to their advantage, but that's not the fault of statistical science. That's people wanting to hear what they want.
Tai_MT
How do you determine what is a "good sample size" other than personal opinion? Even if there were some magical number for a good sample size, how do we know that the method in which they arrived at that conclusion wasn't biased in some way? Just like pollsters calling you on the phone to get their results, they never put into their polls the amount of people who hung up on them or didn't want to participate. They only include those with sad enough lives with nothing better...
Tai_MT
to do than participate in the survey. Statistics are never accurate and are always biased towards whatever it is the study is trying to prove or disprove.
Galenmereth
Statistics are oftentimes flawed or biased, therefore all statistics are flawed or biased.
This is not how it works, Tai. That's making an assumption based on a severely limited sample size (ie. your experience), which is exactly what you're berating statistics for. You don't have to poll the entire human race to figure out whether a substantial amount of people within, say, one country, think or prefer one thing over another. You just need to create a sample size from...
This is not how it works, Tai. That's making an assumption based on a severely limited sample size (ie. your experience), which is exactly what you're berating statistics for. You don't have to poll the entire human race to figure out whether a substantial amount of people within, say, one country, think or prefer one thing over another. You just need to create a sample size from...
Ms Littlefish
Not only do you have to pick the right test to run (a poll is it's own test type and is used at its appropriate time). Pick the wrong one. Result is biased. Bad test.
There are metrics and calculations to determine your sample size, as well as your replications of the test. It's not a static number. There are values and variables your test against, and a way to calculate what your error %, what your outliers are, ect. Do the calculations wrong. Biased. Bad test. Redo it.
There are metrics and calculations to determine your sample size, as well as your replications of the test. It's not a static number. There are values and variables your test against, and a way to calculate what your error %, what your outliers are, ect. Do the calculations wrong. Biased. Bad test. Redo it.
Ms Littlefish
Most of the problem that occurs is at the design of the test and the interpretation of results.
There isn't really much you can other than know the difference and be educated about the design and results of statistical tests. Social sciences are by standard, tricky. Because. People.
There isn't really much you can other than know the difference and be educated about the design and results of statistical tests. Social sciences are by standard, tricky. Because. People.
Galenmereth
a good enough spread, to make up for inconsistencies or area bias. Of course no statistic polling human opinion can ever be 100% factual, but that's not what it's about, either. It's about connecting opinions and numbers and looking for norms.
Statistics, like any study, can be flawed and/or biased. It doesn't mean all statistics are worthless. What about actual factual statistics, like my git statistics? Every time I commit code to my git repository, the computer l...
Statistics, like any study, can be flawed and/or biased. It doesn't mean all statistics are worthless. What about actual factual statistics, like my git statistics? Every time I commit code to my git repository, the computer l...
Galenmereth
ogs the time, date, size of the commit, lines changed etc. I can grab graphs to see when I do most of my work; what day of the week, what hour of the day, etc. Many statistics are like that, too: Pure facts. Don't dismiss everything as one and the same.
Tai_MT
What are "statistical norms" in any case? Those never take into account variables either. You could have a substantially biased result of how much you actually work if you took off a week from work and did absolutely nothing. All those "zero work" days move in and destroy the statistic or alter the actual data in a way that may or may not be relevant, depending on what you're trying to actually track. Everyone always thinks that a statistic can be "factual", bu...
Tai_MT
but always forget about the variables associated with it.
mlogan
Uh no. That's why there are "controls". And any study worth it's salt WOULD try to take into all associated variables. Like, days off of work.
Galenmereth
Tai: I think you need to actually read up on the statistical sciences, because that is basic stuff that is taken into account in any half serious survey... You are right in that people abuse statistics way too much, but there's also a lot of useful and well-researched statistics out there. Like any science, it can be abused for bad, but don't lump it all together...
Ms Littlefish
As with most tests, you are testing a hypothesis. And in statistics a basic test would involve something called a p-value and a null hypothesis. You also have to repeat your test, make controls, determine variables, run the data through tests, plot the data, make fit lines.
There is a lot more to the process then gathering data and noting what percent of people said/did what.
If a test is not accounting for many of the things you have mentioned, bad test. Redo.
There is a lot more to the process then gathering data and noting what percent of people said/did what.
If a test is not accounting for many of the things you have mentioned, bad test. Redo.
Zeriab
Never attribute to malice that which can be adequately explained by stupidity. Then again, don't rule out malice.
Latest Threads
-
-
A way for events to recognize they are close to each other? (5 Viewers)- Started by Kelphy
- Replies: 2
-
-
Latest Profile Posts
Just beat the last of us 2 last night and starting jedi: fallen order right now, both use unreal engine & when I say i knew 80% of jedi's buttons right away because they were the same buttons as TLOU2 its ridiculous, even the same narrow hallway crawl and barely-made-it jump they do. Unreal Engine is just big budget RPG Maker the way they make games nearly identical at its core lol.
Can someone recommend some fun story-heavy RPGs to me? Coming up with good gameplay is a nightmare! I was thinking of making some gameplay platforming-based, but that doesn't work well in RPG form*. I also was thinking of removing battles, but that would be too much like OneShot. I don't even know how to make good puzzles!
one bad plugin combo later and one of my followers is moonwalking off the screen on his own... I didn't even more yet on the new map lol.
