It's good to discuss where this uncertainty come from in the text

I think I wrote that "GiveWell believes...". You could probably steelman GiveWell a bit better than that. Also, you can use `my_variable` to represent variables. It should print it in a monospace font.

Can I use refName? It seems to let me for some but not others. I'd also ideally like to like the numbers such as 90% scarring to defined values that they can be updated more easily but I had trouble making that happen.

This brings tears to my eyes, this is so good!!!!

I don't immediately see why it infers that children get more than one vaccination. Maybe a longer explanation?

You have a weird link here

Try a beta distribution. You can use beta({mean: x, stdev: 0.001}) or something

On first glance, these uncertainties look surprisingly low, but I'd like to check the original source to determine how uncertain these really are

Hopefully I've improved these now but the top two should probably be more uncertain. I just didn't want to change the mean of have the 95% CI spill over 100% can I make a lopsided distribution with its mean at 0.97 but which falls entirely between 0 and 1? That would help me massively with almost all of my distributions.

Try beta({mean: x, stdev: y})

That does make sense! And I agree with you. "Percentage point increases" is a dangerous unit to work with. This could probably be modelled in a different way...

In fact... I might be wrong... but I think percentage points might be dimensionally invalid...

ummmmm.... I'm confused xD

This formula might need a little bit more explanation

can this be on an axis of 0 to 2 or something?

I tried to calculate the uncertainty in the proportion of vaccine preventable deaths cause by diseases targeted by each vaccine by comparing the probability of death numbers in this spreadsheet https://docs.google.com/spreadsheets/d/1Kzv-FJhGDuB59zwQS9JzGmszVXjIj40Y4kUVt-XVED8/edit#gid=1336430465 to the percentage of all deaths numbers in IHME's data on Nigeria. I would just use IHME's data because it was referenced in (I think) GiveWell's report on New Incentives and has upper and lower bounds but: - it doesn't have all the diseases or categorises them too broadly - it doesn't focus on NW Nigeria whereas the first spreadsheet does - minor but it gives "value," "upper," and "lower" numbers for each item and the value isn't the mean of the upper and lower which makes me this the distribution is uneven that chucking the upper and lower bounds into a log normal will be a bit misleading. Now I'm stuck on how to get back the proportion of vaccine preventable deaths from the disease proportion of deaths. `mortality all cause under 5` is the probability of a child under 5 dying in Nigeria. The diseases represent the proportion on deaths of <5's in NW Nigeria caused by that disease. `vaccine preventable disease mortality under five` is the probability of a child dying from a vaccine preventable disease before age 5.

Would be cool to multiplot the proportion of vax-preventable disease deaths attributable to diseases targeted by each vaccine (..._preventable_deaths)

It would be cool if I could insert the mean of the distribution instead of 22.1 pp

look I'd say its negligibly close to impossible for every child in NW Nigeria to be vaccinated >6 times so this this axis might me a tad unreasonable. Can I make it 0 - 100%? I slightly outside of this isn't impossible but this essentially it.

I don't now how to make proportion_on_those_vaccinated_enrolled_in_ni_rct_in_treatment work with refName

Should be fixed

I'd like this to have a number where it say [objectObject] but I've had a fiddle and I'm still not sure what I need to do to make it work.

:( it says ineffective_goods is not defined but it is...

Would be nice to format this record in a way that's a bit easier to read

If you want, you can format x axis to something easier to read, I would recommend in this case formating these as percentages. There is an example in the squiggle tutorial using "tickFormat"

I am also not certain what these variables are. Please specify!

I read the tutorial tickFormat but had trouble implementing it. Does my added explanation make things clear enough that I don't need it or that you could make a specific suggestion for how to use it? I think it's probably better if I leave these as proportions rather than converting everything in the notebook into percentages.

Put a little bit here about your decision-making about choosing these distributions. If you want to be honest about it (It's a best guess based a couple of small things) feel free to put those in.

Why don't you think it's the most appropriate? Maybe put it into an Appendix

An "All mistakes are our own" would be great. As well as maybe: "This notebook is interactive. Feel free to change values/fork or leave comments if you disagree with certain points.

The things in the parenthesis could probably be moved to the section on developmental effects.

Man, I'm just commenting on the stuff I've written at this point xD. If you can, see if you can restructure my questions in a way that match with the headings (Maybe even link to the headings! You can create links to different cells by using the "copy link" functionality you get from clicking on cell options) and then say it's "This multiplied by this multiplied by this" sort of thing.

Also, a quick overview of what you've added would be important, most notably your decision-making process for choosing the uncertainties, which is the major point

Replace this with refName = (x) => `[\`${Object.keys(x)[0]}\`](#${Object.keys(x)[0]})`

We can do real Bayesian updates (It would be a lot of work, put it in the appendix of improvements)

I'm actually genuinely tempted to do an actual Bayesian update here. This is less than ideal

Would be nice to mark that you've done donation_size = 1 on purpose to get cost per dollar

What's Gavi? Why does it matter what other people ar spending on this program?

Sorry! That's my problem. I don't know why this happens and I'm trying to fix this ASAP.

I don't want this to say "costs, costs and costs." I want it to say "usd_cost_counterfactual_full_immunisation_gavi, usd_cost_counterfactual_full_immunisation_government and usd_cost_counterfactual_full_immunisation_gavi"

More text? Where do these estimates of uncertainty come from

Yeah, this goes into insanity at this point

I'm stuck on finding the CI for the goal seek. Good chance the maths I've written here is wrong

I actually haven't the faintest how to integrate these parameters across all age groups... maybe with excel...

I need to find a way of presenting an asymmetric distribution, knowing the 95% CI upper and lower bounds and the mean.

these don't have uncertainties yet because of my debarcle above

Come back and imbed actual values here and link to relevant cells