Whose good at percentages please? I have been trying to work this out for two hours using excel and I am defeated

[Myheadhurtsagain]

Hi, I am doing a Masters degree at the grand old age of 47. It's history based. My history is good, but my maths isn't! My head really hurts and I am truly stuck. Please can someone help me?!

I have to do select a proportion sample of two groups, and I just cannot work out how many in each group I need to select to make it correctly proportional.

I have put a screenshot on, but basically, I have two groups: survivors of a tragedy, and victims. I have to study 30 of the surviving fathers' children, and 90 of the victims' children.

They were separated into two different groups: A and B (which reflected their different shifts), and I have to make sure that my sample groups of children have the correct proportions of their fathers who were in group A, and those who were in group B.

So, survivors: 4 survivors in group A, and 18 in group B. Total 22 survivors.
Victims: 20 in group A and 78 in group B.

I have to study their children.

I need to select 30 of the survivors children and 90 of the victims children. I have not yet found the kids so don't know how many there are.

How many fathers do I need to select from groups A and B for survivors for it to be proportional?
Ditto for victims?

Many thanks

Chat gpt? I’ve thrown all sorts of accounts/business/tax problems at it and it usually comes up with a formula and an answer!

5 survivors children from group A, and 25 from group B, because you can’t have part-children.

Then 18 victims children from group A and 72 from group B.

Although I’m presuming that each father contributes one child - you may want to set a “number of kids per father” if that’s not the case.

I don't think this makes sense as a question. You have an aim to study three times as many children of victims as children of survivors? But you haven't traced any to study? So just get on and find as many as you want but keep the proportions right- 3 times as many children of victims as children of survivors? The number of fathers is a red herring- are they all definitely fathers at all?

Who knew?! Thank you, it has given me the same answer as the pp below

Thank you! I am not allowed any more than two kids per father. Does that change anything?

I agree with @napody you need to keep the ratios the same, 4:18 and 20:78 which is roughly 18%:72% and 20%/80% and then apply this to the specific numbers of 30/90 that you require

So split across

group A as 5:25 people

and group B 18:62 people

Yes defo fathers. I have done the rough overviews so I know those but don't want to overcomplicate it.

My issue is how to keep the proportions of children correct for groups A and B and for survivors and victims,

And I have to do that based on the original number of fathers who died, not their children (because group A may have had twice as many children per man than group B, for example)

Thank you. Does it matter that I am limited to a max of two children per father? (Some of them had up to a dozen children).

I agree, I don't think it makes sense as described.

duh maths after wine not great - should be 18:72 as PP’s have said

If you are limited then presumably you are told which fathers relate to which children otherwise how do you exclude their 3rd 4th or 5th children?

@whoopdeedoo that's part of the problem. I need a random way to select just two of the children where the families were big.

Ah I'm with you now, apologies. And I agree with @youvegotafastcar and @whoopdeedoo who's sorted the numbers in groups A and B.

I'm not told which fathers relate to which children; I have to find them. I think I will have 280 in total and need to study 120 in more depth. But I must keep the proportions of group A and B correct while doing so, along with survivors v victims, while not allowed to use more than two kids per family. It's such a headache working out the numbers.

Do the ones given above work if I have to limit to two kids per family?

So you can see which children are related to each other/same father?

@whoopdeedoo no, the two kids per father limit is set to ensure that my study is not skewed by using eg five kids from four big families. The tutors want me to ensure I choose lots of different families to ensure more representation and to avoid "skewing" the figures with eg a few large families.

So do I use the numbers pp above have (brilliantly) given but just not choose more than two kids per family?

Yes, it doesn’t have to be random, they will be either victim or survivor and you need to select enough to get to the target figures, in the specified ratio, without using more than 2 from same victim/survivor.

@whoopdeedoo thank you. I thought it would not be good to start with surnames in alphabetical order and just count until I get to the right number because the tutors would say "you only got to S surnames, what about all those who started with T-Z?" That's why I thought it had to be randomised.

Thoughts?

I'd simply go back to AI on this again and ask ChatGPT to select your sample at random and screenshot the evidence to show you didn't select the people.

Can you put them in an order and give each one a number, then use a random number generator?

If you pick one that has already had two siblings chosen you could just eliminate them and pick again?

@BreadInCaptivity @Dumbles both ideas I'd had, thank you, I think this is the way forward.

You do have to randomize. The letter your surname starts with can be tied to the culture that you are from. So, choosing alphabetically can skew your results (assuming, of course that your group is culturally homogeneous).

Yes, that’s how I’d do it.

@McSpoot thank you.

I love that there are so many intelligent Mumsnetters who are happy to help me. Thanks everyone! My brain was hurting and I was stressed and worried I wouldn't be able to work out how to do it. I appreciate your help and it means I won't worry about it all night.