Good Morning Les: I had made a minor sort of disclaimer in the original that the stats were not the primary thing in this particular report (in my view). As you know I think, I am far more used to dealing with randomised trial data (RCT) where stats are rather easier to follow (!) and the key issue as you know - is to reduce bias (defined as a deviation away from the "truth" in a systematic direction).
But in truth - the key issue here actually is to reduce bias in this observational retrospective study (You cannot do it any other way of course) as well. I am sure you would agree. That is the bias stemming from using just one source of data. . . and ignoring for example - perfectly valid but more accurate than from one person alone - equivalents of 'eye-witness' accounts. I think the authors' construction of these two alternate lists is remarkably creative and cognisant of the 'modern world' and data from social media. [Thank you for the methods references BTW. I do not know this literature.] Having said all that you are obviously quite right - it is important to drill down into such reports. In fact if you have participated in COVID prevalence studies you are of far more of a 'big-E' Epi background than I! So actually - I submit that you have have better insight into these types of stats-interpetations. Nonetheless, this is my own slightly more analytical take than was in my first posting to the list. Herein on when i use the words of Jamaluddine et al I will italicise them: i) Jamaluddine et al start with the context box background - which could be a "rationale" for the study: " Multiple sources have questioned the accuracy and reliability of mortality figures reported by the Palestinian Ministry of Health (MoH) in the Gaza Strip during the ongoing Israeli military operation. Critics have raised concerns about exaggerated numbers of deaths, citing issues such as potential data manipulation, missing identifications, and inconsistencies in demographic classifications. Others point to potential underreporting, citing deteriorating data collection systems and unretrieved bodies. " ii) Furthermore: " The (Palestinian) MoH published death tolls during one of the previous large-scale Israeli military operations in 2014 were considered reliable, coming within 4% of the UN’s and 8% of the Israeli military’s estimates." However: "quality of MoH mortality data appeared to deteriorate, as indicated by increasing numbers of unidentified (ie, no name or other unique identifiers)3 decedents (table 1, figure 1). The escalation of Israeli military ground operations and attacks on health-care facilities severely disrupted the latter’s ability to record deaths electronically. . . etc " iii) They note that other lists were made up using a variety of differing ways - and they are discrepant with each other. They have thus " combined three data lists—official hospital lists, an MoH survey, and social media obituaries " . (iv) One of your main queries is this I think: "i would like to have seen better discussion of the details of how each of the three lists were constructed." My response: Do you not think that the methods on p.2 under "data Sources" - is pretty explicit? For eg: They start with the - " 22 368 decedents who died in hospital or who were brought to hospital morgues for whom Palestinian ID numbers, names (first name, father’s name, grandfather’s name, family name), age at death, and sex were reported. " They marked what were " aggregate numbers of hospital-reported and media-reported unidentified deaths (n=9692)." The they go about trying to cross-tabulate the unidentified deaths " following how the " MoH retrospectively identified some of these decedents, reducing the cumulative proportion of unidentified deaths " This led Jamaluddin et al to their so-called " first list for capture–recapture analysis (hereafter, the hospital list)." And then a new list from so-called " raw survey data from the MoH and used these as our second capture–recapture list (hereafter, the survey list. . . " And they go on to detail how they used Facebook/ social media etc etc to disseminate the survey. I think they do give a fair bit of detail. How one would *replicate* the study is perhaps what you are after. . .? Indeed that is a key question. Doubtless in my mind, they have a study manual that could be probed for that level of detail to enable a scientifically sceptical person or someone wishing to apply their techniques to say, the war in Mynamar (for instance). As you know a civil inquiry usually gets that added level of detail, and it may be that the peer reviewers have already seen/done that. I have - I fear - no further time right now. But I will (& I assume you will too) look at their point at the end: " Data sharing All the data used in this study are publicly available. All the data and analyses are available on https://github.com/ZeinaJamaluddine/gaza_ mortality_capture_recapture." v) Returning to the point of potential for bias Your summary is that: " but on a strictly Venn diagram POV, the lack of overlap amongst the three lists suggest under-reporting is likely. " I think that this is exactly the point. " we provide an estimate of mortality that accounts for under-reporting. We estimate a total mortality from traumatic injury of 64 260 (95% CI 55 298–78 525), implying that the MoH reports underestimated deaths by 41%. . . . (and) emphasises the importance of leveraging statistical methods to accurately estimate armed conflict-related deaths, rather than solely relying on reported figures. Implications of all the available evidence The high mortality rates shown by our study, combined with previous evidence, underscore the severe crisis in the Gaza Strip. Our findings validate concerns raised by Palestinian and international organisations, including reputable human rights and humanitarian organisations and UN special rapporteurs, about the scale of civilian casualties. " In actual fact the totality of the experience of death from war-related causes in the Gaza strip is likely to be more fully represented by the entire box enclosing the Venn diagram you reproduce. What this analysis has done is to reduce the bias, and come closer to the limits of that box - which might grandiloquently be called - 'the truth'. I do not know if I hit at least some of your concerns. We can continue this nerdy discussion off-line as maybe some would tire of it here. Be very well, H -=-=-=-=-=-=-=-=-=-=-=- Groups.io Links: You receive all messages sent to this group. View/Reply Online (#34650): https://groups.io/g/marxmail/message/34650 Mute This Topic: https://groups.io/mt/110586134/21656 -=-=- POSTING RULES & NOTES #1 YOU MUST clip all extraneous text when replying to a message. #2 This mail-list, like most, is publicly & permanently archived. #3 Subscribe and post under an alias if #2 is a concern. #4 Do not exceed five posts a day. -=-=- Group Owner: marxmail+ow...@groups.io Unsubscribe: https://groups.io/g/marxmail/leave/13617172/21656/1316126222/xyzzy [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
