|
Excel used to not let you change an axis to log scale if some of the data were in the negative range. It may still be the case.
|
# ? Jun 10, 2022 16:04 |
|
|
# ? Jun 6, 2024 19:36 |
|
|
# ? Jun 10, 2022 17:31 |
|
It looks like
|
# ? Jun 10, 2022 17:56 |
|
That looks like a flat, straight, line would fit just as well as that curve.
|
# ? Jun 10, 2022 20:34 |
|
clearly that datapoint at d11 isn't an outlier and needs to be counted
|
# ? Jun 10, 2022 20:35 |
|
N=6. Yeah, that's worth a graph, why wouldn't it be?
|
# ? Jun 10, 2022 20:50 |
|
I count eight points on that plot.
|
# ? Jun 10, 2022 21:39 |
|
ultrafilter posted:I count eight points on that plot. I count nine There's one way up top near the title
|
# ? Jun 10, 2022 22:02 |
|
And it's completely driving the fit of their "model" too.
|
# ? Jun 10, 2022 23:37 |
https://fi.somethingawful.com/safs/titles/1c/42/00422239.0005.png The N is the number of patients, the data point is the mean of all of them which does raise the question what the gently caress happened to one or all of them on that day 10 visit I mean, to be fair it could totally be a legitimate data point, aliasing and all. this is it, for the curious https://investor.cariboubio.com/news-releases/news-release-details/caribou-biosciences-announces-positive-initial-data-cb-010-anti ahahahahaha the n=5 for 'evaluable' so they can claim 100% response rate e: which seems to be because the 6th patient was too new for the trial submission cutoff date Watermelon Daiquiri has a new favorite as of 00:32 on Jun 11, 2022 |
|
# ? Jun 11, 2022 00:17 |
|
poo poo like this really erodes my confidence in the "softer" sciences. I've spoken with biologists that happily put their small samples into a statistical black box that yielded "results" with absolutely no understanding of the process. Charts like the above are just insulting. I hope this poo poo isn't as widespread as it seems.
|
# ? Jun 11, 2022 01:40 |
|
Count Roland posted:poo poo like this really erodes my confidence in the "softer" sciences. I've spoken with biologists that happily put their small samples into a statistical black box that yielded "results" with absolutely no understanding of the process. Charts like the above are just insulting. I hope this poo poo isn't as widespread as it seems. I'm a non-academic research statistician and this is something I spend a fair amount of time thinking about. The tl;dr is that there is reason for concern but not for despair. I want to do a more serious treatment at some point but I still need to spend some time thinking about what exactly I want to say.
|
# ? Jun 11, 2022 02:24 |
How should they proceed with studies that can only have a small sample size out of necessity?
|
|
# ? Jun 11, 2022 03:05 |
|
lol 'softer' sciences
|
# ? Jun 11, 2022 03:07 |
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3134305/ I think this makes a compelling argument for keeping small n studies around In defense of that graph, the trend line does do its job of demonstrating their hypothesis of how the dna is replicated in the body over time in general. Biology by nature is going to be very noisy anyways. Watermelon Daiquiri has a new favorite as of 03:34 on Jun 11, 2022 |
|
# ? Jun 11, 2022 03:28 |
|
You have to do small studies in medicine when dose finding, it would be unethical to go straight to big study sizes. The issue is that PK curve isn't supported by their data.
|
# ? Jun 11, 2022 07:11 |
|
As someone with a background in super soft science (linguistics/phonetics), I was always taught that you should just present the data as is instead of making graphs if your sample size is that low. Just skip the statistics entirely and just look at the data. It's not like your p value or error bars or whatever is meaningful anyway. And then just write "looks like people get less cancer around day 10 in this sample" instead of trying to visualize it
|
# ? Jun 11, 2022 08:32 |
well for this its the concentration of the specific gene edit/creation. It doesn't seem too out there to say 'the mechanisms at play here see a maximum of replication after around 10 days' and you must be aware there are many many people who need a nice neat curve to understand something and for whom a bunch of points is meaningless. Also, we don't know the context for this particular graph. Considering the study sponsor is presenting this at an industry(?) event right now, it seems reasonable they want to make things look pretty to easily get their point across. Heck, as a layperson I could totally see them pointing to this and saying 'well as you can the max concentration of copies is around 10 days in, after which it falls off over a month. Realistically you only need to watch for a CRS for X days' or something like that Watermelon Daiquiri has a new favorite as of 09:42 on Jun 11, 2022 |
|
# ? Jun 11, 2022 09:38 |
|
Yeah understanding the expansion of the cells is really important for understanding toxicity and efficacy, but what they've done is fit what would be expected from a normal autologous CAR T product to their allogeneic one. They can't infer the timing of peak expansion from their data. The expansion of cell therapies is also highly variable and the lack of any kind of range means you can't understand what's going on at day 10. The research and product is pretty interesting, being a potential off the shelf lymphoma cell therapy, and it's gene edited in a way that ~should~ improve the efficacy. However, that graph is bullshit.
|
# ? Jun 11, 2022 10:46 |
|
Regarde Aduck posted:lol 'softer' sciences
|
# ? Jun 11, 2022 14:12 |
|
Count Roland posted:poo poo like this really erodes my confidence in the "softer" sciences. I've spoken with biologists that happily put their small samples into a statistical black box that yielded "results" with absolutely no understanding of the process. Charts like the above are just insulting. I hope this poo poo isn't as widespread as it seems. Since when is biology a “soft science”?
|
# ? Jun 11, 2022 14:49 |
|
M.D.s doing research and leading a study even though they have no experience and just started after medical school
|
# ? Jun 11, 2022 14:52 |
knox_harrington posted:Yeah understanding the expansion of the cells is really important for understanding toxicity and efficacy, but what they've done is fit what would be expected from a normal autologous CAR T product to their allogeneic one. so what you're saying is they are *definitely* pointing to it for investors Whooping Crabs posted:M.D.s doing research and leading a study even though they have no experience and just started after medical school To be fair, checking her CV she did years of a clinical residency to start with. did she lead studies in that??
|
|
# ? Jun 11, 2022 15:14 |
|
Watermelon Daiquiri posted:To be fair, checking her CV she did years of a clinical residency to start with. did she lead studies in that?? Yeah, I was being hyperbolic and unfair to her. Just speaking from experience, I have worked with some very demanding and bad MD researchers (and also some good ones). The graph would not hold up to publication level scrutiny, and that's why it's on a research poster.
|
# ? Jun 11, 2022 15:21 |
|
|
# ? Jun 11, 2022 15:39 |
|
King Hong Kong posted:Since when is biology a “soft science”? Since the gender ratio of biologists got to be about even.
|
# ? Jun 11, 2022 19:55 |
|
If I were rich I would build a machine that fires buckshot at a target and then a robot comes in and draws a trend line on it.
|
# ? Jun 11, 2022 20:03 |
|
I see you've played American connect-the-dots before
|
# ? Jun 11, 2022 20:09 |
|
https://www.theonion.com/gently caress-everything-were-doing-five-blades-1819584036
|
# ? Jun 11, 2022 20:16 |
OwlFancier posted:If I were rich I would build a machine that fires buckshot at a target and then a robot comes in and draws a trend line on it. It's fine we have the R squared on there!
|
|
# ? Jun 11, 2022 20:20 |
|
Probably one of the Onion articles that they were more OK with coming true
|
# ? Jun 11, 2022 20:29 |
maps are charts right?
|
|
# ? Jun 11, 2022 21:35 |
|
Watermelon Daiquiri posted:maps are charts right? I demand Latvia cease hostilities and end this naval blockade at once!
|
# ? Jun 11, 2022 21:41 |
|
Lemniscate Blue posted:Since the gender ratio of biologists got to be about even. I honestly can't tell what joke you're reaching for here.
|
# ? Jun 11, 2022 22:51 |
|
jeebus bob posted:I honestly can't tell what joke you're reaching for here. there's a correlation of a science field having more women and it being classified a "soft" science, and it's not a coincidence AreWeDrunkYet has a new favorite as of 23:03 on Jun 11, 2022 |
# ? Jun 11, 2022 22:59 |
|
It's called for a reason.
|
# ? Jun 11, 2022 23:15 |
|
Way back in the day, SNL made a commercial making fun of the Trac II. They added another blade (how ridiculous!) and called it the Triple Track. The first blade pulls the hair out, the second grabs it and the third sneaks up and slices it. And by that graph, we should be approaching infinite blades. I'm picturing a tank tread like belt with blades.
|
# ? Jun 11, 2022 23:28 |
|
King Hong Kong posted:Since when is biology a “soft science”? It's harder than psychology but softer than chemistry. As I mentioned above I've met/known biology students with no real mathematical understanding-- they just plug numbers into a computer. I've not encountered this in fields like neuroscience or physics though I'm sure it's still possible. This isn't meant as a dig. Lacking math doesn't invalidate a field of research. Seeing scientists gently caress up basic statistics is concerning to me though.
|
# ? Jun 12, 2022 00:42 |
|
It's this but real life: https://pages.cs.wisc.edu/~kovar/hall.html quote:Electron Band Structure In Germanium, My rear end
|
# ? Jun 12, 2022 00:50 |
|
|
# ? Jun 6, 2024 19:36 |
|
does anyone who puts stock in a distinction between hard and soft sciences ever put their discipline on the soft side? anyway, whenever I hear hard science I think of
|
# ? Jun 12, 2022 01:30 |