#20 - Quality control investigator: Upon testing samples

[heartofsunshine]

I chose A on this problem because of the fact it said "sent by our field inspectors from VARIOUS manufacturing locations". No where did you know if this included 100% of the locations or only a small sample size. So if it doesn't include 100% of the locations, this data is invalid, that is why I went with A) bases a conclusion on too small of a sample size.

I also don't really like how the testing was done. Usually you think there would be a more methodical approach to this, where you would take (# of defective items / the total garments from all facilities) to reach your target of 5% or below. Taking random items and doing a random test like this is flawed in itself, which I guess is also why answer choice D works (but also C). I thought D and C both pointed to an overarching theme of lack of an appropriate sample size.

Help in understanding what these test makers want me to?? Thanks!

[Adam Tyson]

Actually, Heart, random sampling is a great way to test for things like defect rates, so long as the sample is representative of the total. We don't need to test 100% of the sites, or 100% of the products, to be fairly certain of our results, so long as the sites chosen and the samples chosen are a good representation of all sites and all products, and random selection of the products would help ensure that.

That's what the author failed to consider here - the possibility that the selection of products by the inspectors was NOT random! What if the inspectors were pulling products that they thought, upon inspection, might be defective? That would skew the sample away from being random and representative.

"Various" sites means they went to more than one or two, and that's good. There is nothing in the stimulus to suggest that the sites chosen were not representative. Just the opposite, really - it suggests that they had some variety, which is good. While it is possible that the sample size was too small, the stimulus does nothing to suggest that, so we cannot say (as answer A does) that the sample WAS too small, only that it MIGHT be. Answer A is just too strong, and assumes facts not provided to us.

TL;DR: A representative sample doesn't need to be 100%, but it does need to be representative, and that includes some degree of randomness in the selection.

[mguitard]

Hi, I was wondering if I had the reasoning correct in why D is the right answer:

The quality control investigator assumes that the field investigators and the supplier both examined the same sample items. So, if this were assumed, he'd overlook the fact that they did not. The field inspectors may have, as stated by D, chosen to test those items they suspected to be defective.

Additionally, I know understand B is incorrect now, however, I cannot seem to put into words why it's incorrect.

[Rachael Wilkenfeld]

It's not that the supplier needed to inspect the items, mguitard, but that the items chosen by the field inspector needed to be representative. Answer choice (B) describes a representative sample---the inspector is not more likely to select a defective one than a non-defective one. Answer choice (D), on the other hand, describes the flaw. The argument overlooks the possibility that the items selected were NOT random. That would mean that the rate of defective products found is not reliable to determine the overall defective rate.

Hope that helps!
Rachael

[supjeremyklein]

summary
the field inspector sent 20 units from the supplier and 4 of them sucked
if we get 2 or more units back that suck, then our supplier blows
so our supplier blows

analysis
just because the field inspector sent us 2 or more units back that suck, doesn't mean our supplier blows

answer choices:
(d) says it's the field inspectors who actually blow here...lol

(a) talks about sample representation being sketch, which definitely raises a good question, since the total number of units being manufactured is unknown. but since the argument lacks the necessary information (total number) for this answer choice to have an impact on the reasoning, it can't be the right answer.

(b) talks about the odds of picking a crappy unit out of the bunch. but if the odds of picking a crappy unit are 50 percent, then it would seem to me that the supplier actually does blow, considering half the units are presumably crappy. so this answer choice feels like the opposite of the objective.

(c) is similar to (a). it's uncertain how many manufacturers are involved here. but this is also a lamer answer choice than (a); it might be the opposite of the objective; even if a few or all manufacturers suck, the responsibility lands on the supplier, and since it's the suppliers job to make sure the products aren't crappy, it would lead me to believe that the supplier blows, which would support the argument, not fight against it.

(e) i don't see how the distribution of visits relates to the information provided in the argument

[queenbee]

Hi
When i read this question, i thought, what if something else impacted the quality of the product sampling, after all, there was an intermediate who picked up the samples and physically sent them to the quality control investigator. It could be that the samples were damaged when they were sent and not in the manufacturing plant. So I selected (B) Would you please let me know why this logic is not correct?

I didn't select (E) because it didn't make a difference if the inspectors when to all or just 3 of the manufacturing sites. Either way, the # of defect needs to be less than 5%. If a single plant is an output where 50% of the product is defective, then that plant would need to be further investigated. Similarly, if a single plant represents 5% of the total product produced, and that specific plant had 100% defective units, then it's something that needed to be investigated.

Would appreciate any help you can offer here.
Thank you

[Robert Carroll]

queenbee,

Your logic is correct, but that's not what answer choice (B) says. Answer choice (B) says that the argument presumes the odds of picking a defective item were 50%. The author never said that! Nor does the author believe that - the author probably believes the defect rate is around 20%, because in the premises they give, that's the defect rate in their sample. The issue, as you noted, is that defects in the sample may overstate defects in the entire group. Answer choice (D) points out one reason why the defect rate in the sample may be inflated.

Answer choice (E) is incorrect and that's the idea - it doesn't matter for the argument how many sites they visited or how often.

Robert Carroll

[zebrowski]

Hi,

What threw me off on test day, and what still throws me off, is that answer choice (B) needs a little helping out.

Field inspectors may very well tend to choose items that they suspect are defective, but it does not mean that these items are defective.

We cannot infer from the fact that they tend to choose items that they suspect are defective to the conclusion that the selection was biased.

This is not a valid inference.

To make it valid, we would have to to add a further premise saying that if field inspectors suspect an item to be defective, it is probably defective.

Without this further premise the argument does not work. For all we know, field inspectors could be incompetent or inept or inexperienced and their hunches could be off the mark.

So we need to add this premise, and yet we are asked not "make assumptions that are by commonsense standards implausible, superfluous, or incompatible with the passage."

I find it hard to tell which assumptions are "implausible, superfluous, or incompatible with the passage" and which are not.

[Jeff Wren]

Hi zebrowski,

I believe you are referring to Answer D rather than Answer B.

I think that you may be approaching this answer and the flaw in the argument the wrong way.

We don't need to prove for certain that the selection process was biased, simply the possibility that it may have been biased creates the flaw, even if we aren't 100% certain of whether there is bias or exactly how much bias there was.

Think of it this way, if you were in change of determining what percentage of the products are defective, how would ideally make this decision? You would probably devise some sort of random number generator to check the products without any human decision making involved at all. As soon as you let people (inspectors or otherwise) choose which products to check, you've added another variable to the experiment that you are not intending to test and thereby invalidating the results.

Notice how Answer D states "overlooks the possibility that...." Simply overlooking the possibility that the inspector may have chosen the products in a way that is not-random (and in fact may very likely create a higher percentage than the overall defective rate) is the flaw.

As for commonsense assumptions, the appropriate assumption here is that, if the inspectors tend to choose products that they suspect are defective, then there is a good reason to suspect that the defective rate would be higher than a that of a random sample. Of course, we don't know exactly what products are being manufactured, but it may very well be the case that some defects are apparent simply by glancing at them, such as a product that is visibly broken/damaged. Don't bend over backwards trying to imagine that the inspectors are completely incompetent when there is no reason make that assumption.

Here's a more extreme example a similarly flawed argument.

Imagine that I'm trying to take a poll for the next political election. However, I take the poll at a rally for a particular political party and my results indicate that 99% of the respondents are all planning to vote for a particular candidate, who happens to be the candidate from the party that the rally is for. I then conclude that this candidate has a 99% chance of winning based on my poll.

Do I know for certain that my sample was biased? I haven't proven it, but commonsense would indicate that the sample was very likely biased based on the manner in which I conducted the poll, which was anything but random.

In short, anytime the sample is anything other than a random/representative sample of the total, that creates a potential for the unrepresentative sample flaw.