Edexcel S3 - Wednesday 25th May AM 2016

Original post by Euclidean

To be honest, aside from some combinations of variables trouble I had yesterday (squaring not squaring etc), I got all of S3 on my first run-through in like 5 hours. Confidence intervals just sort of make sense if you think about the concept.

Error we can assume is normally distributed. It's value depends on how big of a sample we take, so bigger sample means smaller error. The interval depends on what percentage (for 90% this is 5% either side) we don't want to concern ourselves with on either side of the true mean so we can use normal tables to find that corresponding Z value (as error is normally distributed).

That's probably a patchy explanation in terms of the actual statistics but it's how I reasoned with the idea.

Interesting thanks. Someone explained this to me very well:

The way I learnt it was using bottles of wine. Let's say you have 1 massive bottle of wine (worth 12 small bottles in volume), and 12 small, individual bottles of wine. You'd expect on average the small bottles to vary naturally , some are higher than the mean, some are lower than the mean so these differences cancel each other out so it has a smaller variance. Using the formula it would be 12Var(X). However with the large bottle of wine, the variance is amplified as you are taking 1 bottle then multiplying it by 12, using the formula it will be 144Var(X). Hope you can understand it.

Reply 442

username1162972

18

Original post by L'Evil Wolf

its not even that hard though and doesnt even cover much marks in exam.

But it belongs in Science and Statistics

Thats why science and number crunching statistics is ****. I don't have time to learn words.
Geometry will rule the world.

Posted from TSR Mobile

Reply 443

Original post by physicsmaths

Thats why science and number crunching statistics is ****. I don't have time to learn words.
Geometry will rule the world.

Posted from TSR Mobile

Euclidean approves

Reply 444

Original post by physicsmaths

Thats why science and number crunching statistics is ****. I don't have time to learn words.
Geometry will rule the world.

Posted from TSR Mobile

Yeah I agree, statistics after a while becomes the same regurgitated content. Wilcoxon would have been fun to do instead of sampling aha :l

Reply 445

justalawyer1998

5

Is simple random sampling (i.e. random number sampling/lottery sampling) not suitable for when the POPULATION or SAMPLE is large? The book says population, an old mark scheme said sample, baxter's notes say sample - which one is it?!

Reply 446

Original post by Nikhilm

Is simple random sampling (i.e. random number sampling/lottery sampling) not suitable for when the POPULATION or SAMPLE is large? The book says population, an old mark scheme said sample, baxter's notes say sample - which one is it?!

Both, but the sample is more important.

Imagine trying to assign 10,000 people a number with no overlaps. But then imagine trying to sample 1,000 people from that population with random numbers. Pain in the a** if you ask me :lol:

Reply 447

username1162972

18

Can some one explain degrees of freedom. Have no clue what the book is going on about. Ch1-3 and Ch5 are easy. Ch4 is a nightmare.

Posted from TSR Mobile

Reply 448

Original post by physicsmaths

Can some one explain degrees of freedom. Have no clue what the book is going on about. Ch1-3 and Ch5 are easy. Ch4 is a nightmare.

Posted from TSR Mobile

Basically it's just a measure of how much information you need to know before you can work the rest out

With a goodness of fit test (checking if a distribution is a good model) it's v (degrees of freedom) = n - 1 where n is the number of columns (data points) you have, but if you use the data to calculate something unknown like p in a binomial model you have to take another one away so it becomes v = n - 2 or v = n - 3 if you're doing a normal test and need to estimate suitable mu and sigma from data

With contingency tables the degrees of freedom v = (r-1)(c-1) where r and c are the number of rows and columns respectively

Degrees of freedom alongside significance level determine the critical value of the chi square statistic which would lead to a rejection of the null hypothesis, have a look at the formula booklet if you haven't already. You'll see what I mean

(edited 7 years ago)

Reply 449

paradoxequation

2

Anyone know where the past papers start from for our spec? I've printed all the ones from 2007, there weren't any Jan papers were there?

Reply 450

paradoxequation

2

Also, the book literally says that stratified sampling is used when the sample is large. But then in EX2B Q4c, it is a disadvantage of stratified sampling if the sample size is large. Lol, so which is it?

Reply 451

Original post by paradoxequation

Also, the book literally says that stratified sampling is used when the sample is large. But then in EX2B Q4c, it is a disadvantage of stratified sampling if the sample size is large. Lol, so which is it?

The first definitely, The second point doesn't make sense.

Reply 452

Will we be expected to apply continuity corrections in s3 questions?

Reply 453

justalawyer1998

5

Any hard non-edexcel papers? @Zacken

(edited 7 years ago)

Reply 454

justalawyer1998

5

Original post by L'Evil Wolf

Will we be expected to apply continuity corrections in s3 questions?

Not for anything concerning samples because, if n is large, by CLT you just assume its normally distributed, so no correction required. Obvs it would be required for an S2 type q, but that's highly unlikely to come up in S3

Reply 455

Original post by L'Evil Wolf

Will we be expected to apply continuity corrections in s3 questions?

If we combine random variables from different distributions

Posted from TSR Mobile

Reply 456

Original post by Euclidean

If we combine random variables from different distributions

Posted from TSR Mobile

Original post by Nikhilm

Not for anything concerning samples because, if n is large, by CLT you just assume its normally distributed, so no correction required. Obvs it would be required for an S2 type q, but that's highly unlikely to come up in S3

Thanks yes I see. Have you done a edexcel s3 paper where they required this I did a similar one in madas.

Also to how many dp/sf do we give the confidence interval linits too - 2.dp./3s.f. etc?

Reply 457

Original post by Euclidean

If we combine random variables from different distributions

Posted from TSR Mobile

Also is Ex17 in chapter 3, wrong?

It should be 0.0334 in each tail...

Reply 458