zoo data analysis

[itchyshin]

@Mike-Garratt - I got around to doing the zoo data with lnRR (no phylogeny is incorporated yet).

Male surgical

lnRR (life expectancy)

lnHR (risk of death)

Female hormonal

lnRR (life expectancy)

lnHR (risk of death)

Female surgical

lnRR (life expectancy)

lnHR (risk of death)

More to come - now will work on the new analysis of our meta-analysis data

[Mike-Garratt]

[Mike-Garratt]

Hi @itchyshin

After are conversation yesterday I decided to have a play around with the zoo data to get my head around our sex differences analysis, including the absolute difference etc.

This is not a meta-analysis but lets me think about the best way to present data etc.

I think plotting the normal log-response ratio (I have just plotted the ratio in this analysis) and the absolute ratio will together tell us what is going on with the sex differences data and hopefully make it easy to understand for people. In the figure above I have made these plots side by side for the zoo data.

In the first ratio figure you can see that normal females live longer than normal males. When males are sterilized and compared to normal females this difference disappears, when females are sterilized and compared to normal males the difference becomes bigger. When both a sterilized the sex difference is the same as when both are normal.

The second figure is the absolute difference comparison - how big is the sex difference, ignoring whether it is male or female biased. This doesnt change when males are steriilized, basically because all that is happening is males are living longer and this is counteracting the sex difference in lifespan. There is no difference in a paired difference for the absolute difference because the sex difference in survival is not reduced. For both sets of female sterilization (hormonal or surgical) the absolute difference gets bigger, because females normally live longer and this difference is made even bigger with sterilization. This can also be seen in the paired test.

However, when the absolute difference in survival is compared between both normal animals to both sterilized animals, there is no difference. Sterilization leads to a similar increase in survival in both sexes, and this means that there is a similar absolute difference in lifespan when comparing sterilized animals. This last result is different to what you observed in the meta-analysis.

[Mike-Garratt]

I think that this is a really cool result. Just need to find the best way to explain this simply to people. Might be worth trying to explain what we are doing to Fernando and Jeff and see if they understand the analysis.

[Mike-Garratt]

Happy to have another quick chat about this if you thought it useful. Sorry about the essay!

[itchyshin]

@Mike-Garratt - wow - this is great - this gives me a good idea of what I need to do as well. Yes, it is a good idea to inform Jeff and Fernando. Let me first go at the figures we discussed first and then we can probably set a new meeting next week or so. Thanks!

[itchyshin]

@Mike-Garratt - there are some phylogenetic effects (esp, female hormone) but overall I think it is consistent with our meta-analytic data - actually much is coming from non-phylogenetic effects - these values show how much variance in effect sizes (y) is explained by a particular random effect (total = both phylogenetic and non-phylogenetic effect).

Male surgical

Female hormonal

Female surgical

[Mike-Garratt]

Interesting, thanks Shinichi. As we only have one effect size per species I guess the species value will always be high? Regardless, good that its consistent with the meta-analysis. Thanks for working on this, Mike

[itchyshin]

@Mike-Garratt - Yes you are correct and the species effect is potentally confounded with measurement errors and other errors. I can calculate CIs for each effect and see whether they are sig. Anyway, this is good news as we do not really put two papers out there which contradict each other. Both datasets have strengths and weaknesses which you have pointed out before

[itchyshin]

@Mike-Garratt - I re-opened this - things to talk about today

• what analyses to do
• what figures to create (I can redo Fig 5)
• I am confused with some figs in the text (ask some questions)
• what else?

[itchyshin]

Things to do for @itchyshin

• [x] create the sex difference figure with a tree for the zoo data (work with @Jeff-Lemaitre and @mlagisz )
• [x] finish off meta-analyses for zoo data (life expectancy)
• [x] make sure to do the same for the rate of aging
• [ ] put all the analyses and figures into the HTML (work with Fernando)

[itchyshin]

Updating the list after a meeting with @Mike-Garratt

• [x] Contact Johanna to get the fig for Fig 1
• [ ] recreate Fig 1 & 5
• [ ] Analyses pre and post data

[itchyshin]

@Mike-Garratt - some new analyses

There is a clear outlier (male possums)

I will do more and also - try to do figs as you wanted with a phylo tree.

As you see below - all effects are very consistent

Overall effect is as much as 28% (excluding possums - 20%)

Overall

Sex

Type of Castration

Sex x Type

[itchyshin]

Here are risk-of-death analysis @Mike-Garratt

Some of the slopes (risk) seem to have SE too narrow or precision is too high compared to others (we may need to ask Fernando to check?)

These are - "Procavia capensis" (hyrax) "Suricata suricatta" (meerkat)

Overal

Sex

Type of castration

Sex x Type

Correaliton between lifespan and risk of death

[itchyshin]

Things to chat with @Mike-Garratt

• [ ] what to do with the outlier (possum)
• [ ] what to do with the risk of death analysis (some outliers in terms of precision or SE)

Things still to do for @itchyshin

• [ ] pre-post analysis
• [ ] Fig 1 & 5
• [ ] HTLM including everything (need to decide what to put)

[Mike-Garratt]

Hi @itchyshin

Thanks for sending this through! I had noticed the possum outlier as well and think that we should exclude this as it must be a mistake in the database somewhere, so a problem with the zoo data.

I didnt realize the had already got slope data from Fernando. I am not sure about these outliers, I don't see a problem with excluded them all will we need to highlight this in the methods and so it will be good to check with Fernando at some point.

Are you all good with the plan for the figures? If you want to set up a time to chat then let me know, I am pretty free this week.

Thanks for your work on this!

Mike

[itchyshin]

@Mike-Garratt - thanks for your replies.

Let me work on the figs and other analyses next 2 days (I will update you here), and we can set a meeting mid or later in the coming week.

[itchyshin]

@Mike-Garratt - I am sorry for the slow progress (I am putting all the time but some data matching and checking is taking me long). I have found that Elk is not in the tree - in the tree, it is the same as Red Deer so I will get @mlagisz's help to get this added to the tree (I think for the currently fig, it was just deleted). I am hoping that I can get some preliminary fig for Fig 1 and the analysis for Fig 5 done today

[itchyshin]

This needs to be resolved (@mlagisz - we will chat soon)

[Mike-Garratt]

Thanks @itchyshin , sorry sounds like a pain. Let me know if I can help! Mike

[itchyshin]

@Mike-Garratt - nearly there

[itchyshin]

@Mike-Garratt - one of the package (ggtree) seems a bit broken so I cannot recreate what Johanna did but I will ask her for help a bit later - I am moving onto other analyses (abs analyses, fig 5 and pre-post). The image quality seems a bit low - I will look into later (this package seems to have a lot of issues so I cannot really figure out a lot of things for now).

[Mike-Garratt]

Thanks Shinichi, annoying about the image resolution but otherwise it looks amazing! Some of the surgical effects in males are very strong...thanks for all your work and please let me know if I can help!

[itchyshin]

@Mike-Garratt - these really high ones are usually from low sample sized studies- we could add orchard plots like Fig 2 so people know spp estimates are based on very different sample sizes (precisions)

[Mike-Garratt]

Hi @itchyshin , yes that might be good. They do seem like very large increases in life-expectancy. I did look at the original plot of wallabia bicolor, which I think is one of the big ones, and the increase looks biologically plausible. Wallabia_bicolor.pdf

[itchyshin]

@Mike-Garratt with phylogeny - I will do one with absolute values now (but I have to figure out contrast first)

[itchyshin]

Absolute ones - not done stats but they look all the same (chat tomorrow)

[itchyshin]

@Mike-Garratt - I have added all the contrasts for both analyses above

[itchyshin]

pre and post analyses

They are consistent in terms of effect size (seems a bit small) - pre is not sig as small sample size