Collaborate on Terms and Conditions for Invoice Example for Animal Science with Ease Using airSlate SignNow

Terms and conditions for invoice example for Animal science

okay good morning welcome to the 2022 um 2022 annual ticketing lecture in animal science my name is sembo yang associate professor in the department of animal science at the university provider i'm the chair for today's seminar the university manitoba campus are located on the original lens of the anishipur kri ochikri and dakota and danny peepers and on the homeland of the metis nation we respect the treaties that were made on these territories we acknowledge the harms mistakes of their past and we dedicate ourselves to move forward in partnership with the indigenous communities in a split of reconciliation and collaboration and this annual lecture records and the honors doctor stand chained and do the support for the department of animal science and and the agriculture research at the euros manitoba dr chen obtained his bsc and msc and a phd in animal science from the universe manitoba his graduate training was in animals breeding and genetics under the supervision of the late dr bob parker and he was also awarded a doctor of laws from the united university of manitoba in 1991 dr chill returned his family business in hong kong in 1975 and he is currently the executive chairman director of the harvard holdings limited dr chain's many contributions include support of the ticketing and a center for animal rights science research and the national center for livestock and the environment and the ticketing reading room and the recent recent contribution to the education and research in animal science in recognition his generosity and commitment to the education and research the department established his annual stamina in his honor and today we are very happy to have dr uh bill reed wells and dr bervis professor emiliatos in the department of animal science ohio agricultural research development center at the ohio state university received his bsc and msc degrees from purdue university and his ph phd from ohio state ohio state in daily nutrition throughout his career his research has a focus on forage utilization by dairy cow's relationship between the minerals and vitamins and the health of dairy cows they add formulation vulnerability and fresh coc nutrition and dr wales has also called us two books uh 13 book chapters 143 peer reviewed articles 1 104 and abstracts and the 262 popular plays praise the proceedings articles and his paper has been set all 7500 times and moreover he has the 20 papers that have been sent over 50 times particularly documenting his railroad exporting in the energy vitamin e as learning nutrition he has delivered over 285 invited national and international talks he has received many awards based on national and international excellence including three research awards from the american theory science association accumulating in being named a adsl fellow and dr west's presentation today is incorporating variation into that formula formulation for dairy cows please join me in welcoming dr liz yeah the floor here thank you very much uh it's quite an honor to for this invitation um i will let me just share my screen here and then does everything look okay yeah okay um i i visited the university of manitoba my first year as an assistant professor at north dakota state university that was my first job and i was looking forward to visiting your campus again because it's been 35 years but with covid we we've all learned to be very flexible but again thank you for the honor of inviting me for this uh seminar what i'm going to talk about today is something we started right at the end of my career i'm wishing now we would have started this research earlier because it it it's quite an interest to me and i thought hopefully i'll make it an interest to you um and that's basically how do we use variation uh in feed and animals when we formulate diets um what we typically do let me just give my pointer when we formulate a diet for dairy cows and this is the same um well let me again i'm ahead of myself what i'm going to try and talk about today is first of all understanding variation and specifically understanding sources of variation i want to talk about how we can get some estimates of variation and then part of what i'm going to talk about can be applied today there's there's ways we can use variation today when you formulate diets but there's also a lot of research that needs to be done so for the the students in the audience i'm hoping i piqued some interest in that we can develop research protocols to start using variation better in the future and these these topics will be interspersed they won't be discrete categories but when we formulate a diet for for dairy cows and this is the same for beef i i don't i'm not familiar with swine and poultry but i think it's probably similar for all species we'll sample the feeds and often for dairy all we sample is the forages and we we send it to a lab and get a number or you might be a nutritionist and you're collating corn silage for your region so you have the average for your corn silage or whatever or very often you just go to the the book a table or the software and get get the standard and composition data for a cow for the cow part uh depending on what you're doing you might if you're a research you might enter data for a single cow if you're formulating diets uh in the real world you might take the the pin mean the average body weight average milk production average milk composition for a pen or if you're feeding a single a single diet to the whole farm you're going to use farm means and then basically you're going to go to whatever software you choose it's going to use linear programming to get a diet that meets those requirements the diet is based totally on either single samples or means the cow is going to be based or the cow data is going to be based on means everything is means and there's no no no place in any of this software and the software i'm familiar with where it says what's the variation in the composition or what's the variation in the requirements and so i'm going to talk about how it's important that you start considering variation in both feed or diet and the cow and again some of this will be future things you need to think about in the future some of it will be things you can use today okay so we'll start with the feed and then we'll move to diets and then to cows so why is standard deviation or variation in feed nutrient composition important well first and this can be it should be used today is variation in nutrient composition affects economic value a highly variable feed is worth less than a consistent feed even if the average composition is the same a variable feed should be discounted i'll go into more detail on this in a bit we can use uh variation and nutrient composition of feeds to to produce what i call semi-quantitative safety factors and this is semi-quantitative because there's still still some things we need to learn i'll give some suggestions but these should be be adjusted as we get better data and then lastly um poultry and swine or at least poultry and maybe swine use some some some nutritionist use what's called stoichastic formulation rather than simple linear programming this is to my knowledge not used in dairy at all i'm going to talk briefly about this but i think this has great potential and down the road as we generate better data this may become the standard right now it is not used in dairy this is most of the histograms i'm going to show uh an actual data i'm going to show in this seminar is real it's data we've collected these are not made up histograms this is the crude protein composition of two sources of distiller grains uh a and c i took b out just to make it a little less cluttered a is highly variable c is a fairly consistent uh source relative to crude protein on average they are essentially the same they both have essentially 30 protein so on average they're equal and i'm going to assume that fat and and fiber and so on energy are all on average the same but on average they're the same but economically you should pay substantially less for for distillery feed a distiller grains a because it is variable more variable and the reason it's worth less is there's going to be a lot of times you're going to get a load of distillers that's quite low in crude protein and if you don't do anything you just keep using 30 percent as the mean those days or weeks when you use this low low protein distillers the diet will be deficient in protein and and theoretically you should lose milk on the other hand when it's high in crude protein you get no benefit because you decided that 30 percent meets the requirements of these cows so feeding more protein has no value feeding less protein loses milk so you lose money and feeding more protein than what they need is a waste so there's one reason the highly variable feed is is worthless is it increases the risk for deficiencies and lost milk the option then is okay every load of distillers i'm going to get i'm going to analyze it and reformulate my diets and that will prevent the these deficiencies but there's a cost there's a cost to sample there's a cost to analyze and there's a cost to reformulate you don't need to do that if you're feeding distiller c if you're feeding a you'd have to add this cost in or lastly you might say well i don't want to do all this reformulation i don't want deficiencies so i'm just going to feed a higher protein diet i'm going to over feed crude protein which is is a wise decision but it's still a cost over feeding usually costs less than deficiencies but this costs more than if you fed distiller seed so high variability reduces the economic value of a feed and it should be fed or sold at a discount or purchased at a discount um this is a again real farm data we sampled several farms in an experiment we did we went out not every day but several days measured in this this histogram i'm showing crude protein excuse me um and there's two farms here one is is what i'm going to call a fairly consistent farm that's the red one and then the green one is a more variable farm on average both these farms formulated to the same crude protein but this uh green farm the more variable farm it's going to have a lot more days where there's an increased risk of being deficient and if i was the nutritionist on the green farm and i and i couldn't work on reducing variation which should be the first goal but let's say i couldn't do that if i was the nutritionist i'm going to say i've got to prevent these deficiencies i don't want to lose milk so on this grain farm i can't feed the same level of crude protein or calcium or whatever nutrient you want i have to feed more i have to over supplement to prevent deficiencies and that's exactly what what they do um and this these are just examples here um for the for the green farm it's the same basic cows everything's about the same this nutritionist said okay i'm going to formulate to 17 rather than 16 and equalize the risk of deficiencies and and that's the wise decision but it increases the waste of protein which is an economic cost and it's an environmental cost so this is what a lot of people do they just say i'm going to be feeding variable feeds i'm just going to over supplement and and this is better than not over supplementing because you want to prevent deficiencies but this is not the the wisest thing to do a much better approach and this is something you can do today is to discount feeds base the the over feeding on on the specific feed we came up with this based on simulation data and first of all there's a few this is why i'm calling this semi-quantitative because i i start this bullet with for variable feeds and you should immediately ask well what's a variable feed and the question is i'm not sure this things like soybean meal corn canola meal would not be considered variable so you could look at the standard deviations for those feeds as a coefficient of variation and maybe if you are feeding feeds that have substantially more variation than that you should apply this approach but not for these these very consistent feeds and i'll go in a little more detail in a bit but for fees that you consider variable you should discount the protein and macro minerals nothing else i want to be this is not discounting energy you're not discounting or changing starch or fiber or the micronutrients just protein and macro minerals and the discount we've come up with and again this needs additional research because this is all simulation is you take 1.8 standard deviations times the proportion of nutrient that feed is going to supply and in this case you're formulating a diet so you don't know exactly but you might say i think uh distillers in this example will provide about a fourth of the protein or or whatever you you think it's going to approximately supply so in this example the distillers had 30 crude protein it had a standard deviation of four and i said it's going to provide about a fourth 25 percent of diet protein so 4 here times 1.8 which is here times 25 of the diet protein and so what i would enter when i'm formulating the diet is for for the the crude protein concentration of distillers i don't type in 30 i type in 30 minus this which is happens to be 1.8 so at the formulation stage my distillers won't have 30 percent it will have 28.2 crude protein and this will adjust for this variation and it's it's much more quantitative than adjusting the entire diet you will feed elevated protein because this on average the distillers has more than 28 but it's discounting the specific variable feed rather than the whole diet and it takes into account inclusion rate now we limit this to protein and macro minerals because the other nutrients there there's risk of both overfeeding and underfeeding protein within reason there's no it's not over feeding protein within reason is not going to hurt the cow it's an economic cost but it's not going to hurt the cow over feeding macro minerals within reason is not going to hurt the cow but for things like ndf there's risk at too high of ndf you hurt intake and milk but at too low of ndf you increase risk for acidosis so the nutritionist would have to decide if i have if i really want to prevent acidosis i might reduce ndf by this this using this equation in the feed but if you're really worried about lower milk production you might actually increase the ndf by this this amount or you might decide i want equal risk and not make any adjustment so for nutrients other than protein and macro minerals the nutritionist has to decide what risk he's willing to take but this is something you can do right now on and this gives you an idea of what the economic value is a variation in specific feeds stochastic formulation which again is not used today in dairy but right now what we assume with the programs we use we set risk at 50 percent because we use means so if you formulate a diet for 16 crude protein half the time the diet is going to have less than 16 percent and half the time it's going to have more so you're saying you're willing to accept a risk of 50 percent with stochastic programming you decide what risk you want this is the the nutritionist decides this and in this example i i might say well i don't want protein to be deficient so 80 percent of the time i want this diet to have at least 16 percent the this would be set by the user this risk is set by the user as is the the specification and then with storycasting formulation the user picks the risk and it's the calculated diet is based on the variation in nutrient composition of the ingredients so it will say you know based on these variations these standard deviations this is the diet you need to feed to have an 80 chance that it has at least 16 and again this isn't used today in in dairy i think it has great potential if we can get good standard deviations so that brings us to this statement here all these things i've said i think are great and wonderful but for you to use any of it you need good estimates of standard deviations i'm going to plug it put a plug in for the new nrc which is now called nascm dairy nrc it has a feed composition uh library which is very very good we spent a tremendous amount of time working on that that library and one thing we really emphasized or tried to do is to get accurate standard deviations so i'm going to say i think right now some of the best standard deviations on nutrient composition is found in that book corn silage there's about a half a million uh observations screened observations in the data set it has a mean of ndf of say 41 standard deviation of 4.8 so you have a good estimate of standard deviation but that's a population standard deviation so and that's not what you want because you got to think okay why why would why would these half a million corn silages why would the ndf vary that's the the question well there's farm to farm there's hybrids you can think of hundreds of things that differ on between farms why the by the corn silage would would differ there's a big one here and that's sampling two cents the feed may not vary but samples can vary because of poorest technique poor sampling protocols lab the lab if you analyze the the same ground if you take 10 samples out of a jar of ground corn silage you're not going to get 10 same answers they're going to be a little bit different that's analytical variation then there's this time factor and it could be low to load field to field cutting season etc so there's lots of reasons feeds can vary when we look at the diet which is really what we care about the diet can vary because of variation in the feeds it can vary because of sampling and analytical and tmrs are extraordinarily difficult to sample so you have a huge amount of sampling variation also tmrs vary based on how you make it and how it's mixed which if it's done correctly tmrs reduce day-to-day variation if it's done incorrectly it increases it and i'll go into that and then lastly you have the cow factor here that they can eat what they want which we haven't figured out how to include that variation we did a large study a few years ago from 50 farms in the u.s across the u.s these were not just midwestern farms uh we we i won't go into all the details but we had numerous feed samples taken and we partition the variation in nutrient composition uh in for farm and sampling and analytical and time or month is what we used and what we found for for forages and all wet products wet by-products farm was significant source of variation and what that means is you cannot or you should not use the population standard deviation so if you don't want to go to nrc and and dig up the the standard deviations for corn silage because a lot of that standard deviation or that variation is farm to farm and if you're feeding a farm you don't care that another farm is different you only care about the farm you're feeding for several concentrates this is soybean meal this is dry corn gluten feed we had canola soy whole cotton seeds that farm wasn't significant use the standard deviation and in fact you can actually use the nutrient composition in good good tables and nutrients you don't need to spend time and money sampling these feeds and then lastly we have dry distiller grains it's in its separate category because it often was it wasn't consistent but very often farmed a farm had an effect so on distillers you use the standard deviations in books with with caution because it there is at least a substantial amount of farm-to-farm variation we also found then is because most farms are not going to have enough samples to get good estimates of standard deviations you might i'm going to encourage nutritionists to sample frequently enough and can collate this data to get good good estimates but if you can't do that we came up with a a way to approximate standard for on-farm standard deviations and this was based on this paper we published several years ago what we found was that 45 on average 45 percent of the variation in alfalfa silage was farm 65 percent of variation in corn silage was farm and for a blend of or a mix or for a variety of dry concentrates it was about 25 of the variation so if you need a a local standard deviation and that's the standard deviation you should use for a farm an approximation is to take the the population standard deviation so you you go to nrc and look up corn silage and find the standard deviation square it because we're working with variances so you'd square that standard deviation and then if it was uh alfalfa you'd multiply that squared standard deviation by 0.55 then take the square root of that and that would be your best estimate of the standard deviation for for alfalfa silage that you're feeding for corn silage you'd multiply it by 0.35 and for concentrates 0.75 so that that's again the best we can do right now okay i want to talk about again back to sources this is real farm data this is from a farm that was feeding bayleads and there's six bales and we sampled these six bales and these are the ndfs we measured ranging from a low of 37 up to 50. you can say well that's really really variable and it is but you need to think how are you going to feed these bales if for example on monday the farmer took these three bales put them into the mixer wagon and made a tmr with you know corn silage etc the variation in these three bales is huge it's 6.8 the average is 44.7 percent ndm with huge variation but to a cow this variation is absolutely meaningless because a cow is not going to just eat this bale or this bail or this bale it's going to eat a blend of all these bales this in this example this is sampling variation and it's meaningless to the cow what matters to the cow is day to day so these three bales were fed today these three bales were fed tomorrow the diet hardly changed 44.7 44.3 that's a trivial change so sampling variation was huge but day-to-day variation was essentially nil so it's important to separate sampling from from uh actual variation if he only fed one bale a day then these are the variations that matter so sam it it's really important when you think about how the variation in samples is it real or is it sampling um and again i'm not going to go in the details here but we did a total variation is farm to farm plus sampling plus analytical plus what we call true day-to-day and you can can partition this if you take enough replicas so farm to farm you sample multiple farms sampling you take multiple samples per per time analytical you analyze things in duplicate or triplican and for day to day is basically everything left over so we can measure all this because if you replicate things enough and then what's left is called true day-to-day and we did this for for several things but we concentrated on corn silage and hay crop silage and most of our hay crop silage is alfalfa or alfalfa grass mixtures and over a 12 month period we we partitioned all this and for corn silage if you look at this is the sampling and lab variation this is what we would say is true month to month variation and for fiber it's the same sampling variation is as big as month to month variation for corn silage so that's saying you know when you get two two samples you're you're as likely and if the numbers from those two samples are different it's as likely to be caused by sampling as it is by a true change in nutrient composition dry matter is is more there's more uh month to month variation in dry matter that makes sense because these are mostly bunkers and and it rains once in a while so on these feeds where there's a high sampling uh variation relative to true variation it's saying sample take duplicate samples but you don't have to sample as often for alfalfa there was substantial day or month to month variation relative to sampling variation this is telling you you need to sample these feeds more frequently but you don't necessarily need duplicate samples and that's pretty much what we we came to to recommend and you need to think here though is when when you send samples in you're going to get data back and you have to think of this as sample data not feed data it doesn't necessarily represent the feed it represents the sample and very often you send a feed sample in you get data it's different from what you use to formulate the diet but in reality the feed didn't change the sample did but the feed didn't but nutritionists see this new number so they go go to their computer reformulate the diet in this case it's totally unnecessary so you're paying for for time to reformulate when you didn't have to and there's a high risk that the the new formula is is bad you know it will hurt milk production so you need to make sure uh when when the sample data change you need to make sure have high confidence that that represents a true change in the feed and the best way to do that is you take duplicate samples and then the average of a duplicate is less likely to have sampling variation than a single sample and a data i'm not going to show when we took duplicate samples we reduced the risk of bad reformulation by about 95 percent compared to single samples so we almost eliminated that risk on the other hand you you if if the sample data changed and you say well i don't believe it i'm just going to leave things as it is but if it really did change if the feed really did change if you feed that diet for too long you're going to lose milk or have other issues so it's really important to know don't don't ignore a change in in composition but make sure or increase your confidence level by duplicate samples that it really did the feed really did change and just some examples here i'm just using protein and starch say the feed has less crude protein than what your sample came back which happens about 16 percent of the time i'm just going to say that happens about 16 percent of the time so if you reformulate based on the new or if you don't if you reformulate based on the new sample data your diet is not going to have enough crude protein you will lose milk on the other hand if the feed actually has more protein than what your sample had you over formulate protein you waste feed dollars for other nutrients it might not be a be a big economic direct economic for example on starch of corn silage say if the feed has more starch than what your sample said you could over feed starch have milk fat depression sick cows and if if it has less starch again less energy less milk income so the bottom line is there is a huge cost to bad sampling and bad sampling or inadequate sampling because it can result in bad diets that have economic and health risk for the for the farm so make sure you have good samples uh i won't go into the protocols but the key to sampling is you want the particles in this sample the particles and it's important you don't sample feeds you sample particles you want this blend of particles to represent the all this if you just go in and you should never sample from a face of a silo because of health risk or a risk of dying but if you just go in and grab a handful here and a handful here and here and here that's the only particles you're going to represent in this bag whereas if you sample all this stuff that he's knocked down these samples might represent this entire face so we've we've written if you're interested in this you've written a couple preceding papers that go into sampling protocols but you it's worth taking 10 or 20 minutes to get a good sample so what we came up with is for feeds that have high sampling error and this is corn silages is the preeminent one you need duplicate independent samples if you're going to sample feed take two of them make sure they're independent and but you only have to sample it lately about once a month or even less if you're working with feeds with high time variation this is day to day or month to month and this is more more likely the hay crop forages you need to sample more frequently week to month but you don't necessarily need duplicates especially if you're taking good good following good protocols thing i want to move into now is tmrs there's a sampling of tmr has high sampling error high risk but there's also significant risk if it varies too much day to day because this is what the cow gets it doesn't eat just corn silage it eats a tmr in most situations and so first let's go backwards and and that is why if you if you go to a farm and you sample a tmr today and sample it again tomorrow why why would it differ well there is feeder mixture variation or error this is largely controllable by good training uh um of of and and good mixer protocols so this is is is controllable largely it takes some work but it's controllable the composition of feeds is controllable or may i should be using the word manageable it means you need to sample these feeds adequately and appropriately but then there's this big chunk of observer variation and this is what the cow doesn't see this this is what you see the lab will analyze things get a little bit different results and again sampling is huge so you control this and you try to you know multiple samples duplicate samples and you try to get good estimates of this feeding a tmr a well and a key key here is well made a tmr will be less variable in nutrient composition than the ingredients used to make the tmr here's an example i put a very simple tmr together 35 corn silage 25 alfalfa 40 concentrate these are the ndf concentrations the average of these feeds this is for the diet this is the standard deviation for ndf for these feeds if i wait take an average weighted standard deviation and adjust for you know it's a variance and do all the stuff correctly the variation in the tmr is 3.1 on on on ndf but this is wrong because this assumes independ dependency that these these feeds vary in in unison they do not corn silage today may be have above average ndf but the alpha alpha might have below average ndf so they these are independent and that independence means that the variation of the of the mix is going to be less than the variation the expected variation we have software that generates these these expected variations based on feed feed variation and we'd only expect with these feeds the tmr should have an s standard deviation of 1.8 which is pretty good so high variation feeds can be used in a well-made tmr and it will reduce the overall variance because of independence and again assuming it's well made this is not an example of a well-made tmr this is not someone did not come in here and put this pay on the bunk this came out of the mixer wagon this way obviously this is going to be highly variable and worse than any of the ingredients so you have to make a good tmr and again with training good equipment a good pmr is obtainable so another question we addressed here is okay how much variation can a cow take and so we conducted a series of experiments this is just one i'm showing with ndf we had three treatments and the treatments here was variation most nutrition experiments the treatment is a mean in other you test 14 protein 16 and 18. we're testing variation we had three three treatments the blue here is the constant and you're gonna see well it bounces up and down but this is you don't know how much trouble we went to to get it this constant we sampled feeds every day we analyzed it every day and we reformulated every day so this is as good this is better than anything you will ever see in the field the average was 25 forage ndf with a little little variation the red bars are what we call the random variation treatment before the experiment we used a random number table and generated these things and and basically what it says say is on monday the diet is going to have 27 percent forage ndf on tuesday it's going to be 21. on wednesday it's going to be 22. but this was a random pattern that all cows on that treatment were fed and the green is a cyclic treatment so we had five days of 26 then we switched five days at 24 5 days of about 28 5 days of 22. so a cyclic pattern this might mimic someone reformulating uh incorrectly but reformulating the means we did not test we were not testing the the effect of 25 forged ndf they all had the same we tested standard deviations the control had uh point that's about 0.7 and again it's very consistent the other two were formulated have the same variance at about 0.24 but remember one is random day-to-day variation the other is a cyclic uh variation so variation it was four times more very more variable in these two treatments four times on average they were the same and surprise i've conducted research for 35 years and this is the experiment that really really surprised me this is not what i expected and what we found was absolutely nothing intakes 25 kilos no effect at all of treatment we've had fed these treatments for 21 days milk these were this was you know 10 years ago so these are pretty high milk productions 40 to 43 kilos a day no effective treatment fat 1.5 kilos a day protein 1.2 kilos a day absolutely no effects on milk production and production measures the only thing we found was that for these treatments the day-to-day variation in intake and milk was statistically greater than the day-to-day variation in the control so cows responded to the diets but it wasn't it was so short term four or five days at at most that that it didn't matter that how the cow could could compensate so variation as long as on average you're about right over a short period of time cows can handle a lot of variation okay um so that's what we've been doing and now i'm going to switch gears and this is where we don't have a lot of data and that is the variation in requirements nutrient requirements can vary because of genetics and you can have two cows producing the same milk but are different genetically they should may have different requirements cows within a pen you might have cows producing 35 kilos or cows producing 50 kilos within a pen those requirements vary day to day it might be hot today cold tomorrow so on so we'd expect some variation requirements day-to-day and then there's a huge for some nutrients there is a huge measurement error in obtaining the requirements most systems software system for for dairy cattle still base requirements on the factorial system maintenance requirement milk lactation growth fetal or pregnancy they sum this they divide by either an efficiency number if if this was energy this would be the me requirement divided by an efficiency factor to get the nel or uh and this or an absorption coefficient if it's minerals and there's variation in all these and there's variation in these we don't know a lot of the variations but we know it's there i'm going to use the magnesium as an example here the ma the new magnesium requirement is a function for maintenance is intake times 0.3 we can measure intake on a farm or in research and so the only variation in this would be measurement error you know you can measure intake every day if you want the 0.3 had a standard deviation of 0.05 which is pretty big but we do know there's a lot of measurement error in this measuring maintenance requirements is prone is very very error prone but there is there has to be some cow to cal variation but we don't know what it is but the variation the overall variation is high on the lactation requirement it's milk which again we can measure on farm or in research we can measure it every day so the only variation here is measurement error the average uh concentration of magnesium in milk is 0.11 grams per kilo that has a fair amount of variation this is probably mostly cow variation not measure we can measure magnesium pretty well but some of it is measurement but most of this is cal so we have reasonably good estimates of the variation for the lactation requirement we don't have very good estimates on the variation in the cow variation in maintenance and again using magnesium as an example the average absorption coefficient for magnesium for total diet is 30 percent it has there's been enough studies we can get a good estimate of the variation in absorption of 0.16 so 50 the coefficient of variation here is more than 50 percent huge but with magnesium we know some of the factors that affect absorption dietary potassium is a big one source of magnesium whether it's magnesium oxide or or feed or whatever and feeding fatty acids or the fatty acid concentration of the diet so we can remove a lot of these sources and include equations to account for this variation and that gets the the standard deviation for absorption down from 0.16 to 0.05 still variable and some of this is is measurement but if we if we really know a lot we can reduce the the estimate of variation in availability coefficients but again the we we need to know this for for some minerals we have no idea what the variation is and especially the trace minerals we have no idea how variable that is net energy high measurement error maintenance energy is a high measurement diet does affect the maintenance effect maintenance requirement we don't know how exactly and of course there's got to be genetics here some cows have to have a lower maintenance requirement than other ones the same body weight there just has to be uh maria in 2015 published a paper and this was a reevaluation of old data but he came up with a mean uh maintenance requirement of 0.38 megajoules per kilogram metabolic body weight but look at the variation the 95 confidence interval so if with a 650 kilogram cow her maintenance requirement could range from 35 megajoules a day to 68. i'm used to calories so eight to 16 mega calories twice huge huge variation the range here has the net energy equivalence of 12 kilograms of milk so there's huge variation in in requirements and again a lot of this is measurement error but some of it has to be cal here are different papers showing these these are calorie coefficients not megajoule coefficients but you can see on one the low on one one experiment was 0.085 times metabolic body weight up to 0.16 times metabolic body weight when we were doing the new nrc we we pit went through all this and put in 0.1 because it's pretty much the average but we also discuss variation and we show the variation in this but remember some of that variation in fact probably a substantial amount of that variation is measurement so this probably is the best overall mean but there is error or true variation now what we don't know and that is covariance among requirements for example i'm using calcium as an example if a cow happens to have an above average calcium maintenance her maintenance requirement is above average above average of what we would calculate does does her lactate does her milk also have higher than average calcium or lower than average or is it not related we have no idea and if it's if they're correlated that could either reduce the total variation in the requirement and that in the total requirement or it could increase it if they're negatively correlated it would reduce the total variation in the requirement if they're positively correlated it would increase it but we have no idea on covariance among these different requirements and then is the absorption or efficiency factor which we use is it correlated to requirements the absorption coefficient almost definitely is i can just the way we calculate mineral maintenance requirements it's an endogenous fecal requirement and that's going to affect the obser the estimated absorption coefficient so for minerals this correlation goes in opposite directions actually going to reduce the effect of variation but for other men for other nutrients we have no idea so the the question some of you should hopefully address in your career and that is how variable are total nutrient requirements you know how variable is the m e requirement for a cow how a of a specific cow if you put in a couch 650 kilograms milk and 35 kilograms of milk what's the me requirement plus or minus what we don't know same for calcium copper etc we have no idea on the variation in the total dietary requirement which is why we're stuck with using the means right now so i want to end here on on some recommendations if you're formulating today what can you do we know things vary what can you do for minerals if you take the requirement i'm going to say let's use the nrc the new nrc requirement i think that's the best currently available and i'm biased it's about a 20 percent we assume if we base it on other species if you take the requirement multiply by 1.2 and formulate to that you will meet the requirements of most of the cows in that group from mp this was based on work we did a few years ago you're formulating a diet you take the average actual milk for the pen let's say it's 40. and the pen standard deviation for milk is 2 or whatever it is you would want your maximum actual the maximum allowable milk to be actual milk plus one standard deviation so on on software now it gives you an mp allowable milk that should be one standard deviation above the the actual milk for that pen the maximum over feeding right now protein is extremely high milk is at least in the u.s is is fairly reasonably priced but still not as high as feed costs so you'd probably want to be less than this maximum and all these came up about through simulation requirements for energy you just feed to the average cow that's because on average energy if the average cow in a pen is over feed energy she's going to put on body condition that she lost earlier when she was under fed energy and then used body condition score to moderate this but minerals 1.2 minerals and vitamins 1.2 mp a maximum of one times the standard deviation or one standard deviation in milk yield of the pen and energy you formulate to the average cow so in summary here control what you can cows can handle variation but control what you can make good tmrs sample correctly etc control what you can but remember observed variation is not the same as real variation observed variation includes lab variation sampling variation so make sure you can separate those out things might not be as variable as you think they are and then lastly in formulation these this over feeding you should never feed to the average uh a cow but the the degree of over feeding should be based on expected variation and the more variation you you control whether it's be grouping similar cows or or good feed sampling and protocols the less you have to over feed but you still always over feed the mean but as you control variation you over feed less and with that i i thank you for your attention and uh i'll be happy to try and answer any questions if there are any thank you very much dr bills for the excellent presentation so right now is for question so you can type your question through the q and a doctor ways we get the first question at the comments from dr holman kashan i got the question i'm reading it right now do you want me to read it or is every can everyone see the question probably yeah you can read okay yeah you can read the question may be good okay i'm gonna i'm gonna summarize it a little bit but he said looking at this data uh i just imagine how often how other often overlooked sources of variation such as cow genetics variation the rumen microbiome could affect feed efficiency performance economics and where do you see the the future precision farming heading and how can we integrate more of this into diet and farm management that's exactly we need to do this if we're really going to get into precision farming we need to start looking at how variable cows are the requirements and the utilization of nutrients by thailand um the way we get estimates of this is is to do better jobs reporting it and and people need to start incorporating or trying to summarize all this variation i like for example i'm i'm going to take some simple things we have hundreds and hundreds of of digestibility data which can reflect the microbiome somewhat let's start with how variable is is if i feed the same diet to a cow how variable is digestible the digestibility of energy we don't know but it could be done with what with the data handling systems we have now that could be done so i think people need to start we need to and and this is being done now you know we publish papers with means and standard deviations but we have all this individual data which is never in the past was never made available now at least in the u.s and i'm assuming in canada if you're getting public grants you have to make all the data available and so i think with this this larger data sets with individual cal data and so on people smarter than me in statistics can start start looking at the variation caused by cow and that might lead eventually to more precise precise diets so right now i don't other than data mining and uh people making uh all the data available so people can look at that that's my only suggestion right now yeah doctor whis i have one question for you about the the requirement variation you put a 1.2 times the requirement of the minerals just but just 1.0 time for the energy requirement so do you think the evaluation the energy is not important to consider for the data formulation now that that uh what i'm saying here is if i feed the average cow 20 again the average cow 20 percent more mineral than she needs that's that's what i'm i'm saying here so i i put on that i'm going to feed that average cow 20 more than she needs that should meet um assuming cows are like like rats basically but if they're like rats and have the same variation that means almost every cow on that pin will be adequate they'll be the most cows will actually be over fed but they'll be very few cows are underfed with energy half the cows are going to be overfed half are going to be underfed with minerals if i over feed by 20 that's has no risk there that's not going to hurt anybody but on energy you need if i fed 20 percent more energy than the average cow needs that means most cows in that pen are going to get gained condition they're going to use energy and that's going to mean at the end of a long enough period the average cow in that pin will be over conditioned fat whereas if i feed to requirement half the cows are going to be underfed they will be losing condition but eventually their milk yield will drop and now they're going to start gaining the condition back so that's why energy i still say you know you need to make adjustments based on condition score but the average cow should be fed the right amount of energy to avoid fit too thin or too fat whereas minerals over feeding because the either the excess mineral won't be absorbed because of down regulation it'll be excreted in urine or it may be retained but for macro minerals there's very little risks that that's their reason it's not saying this is less variable it's just saying there's more risk if you consistently overfeed energy yeah thank you very much yeah you've got more questions okay from dr marcus cadello very interesting presentation could you comment on the potential of the analytical techniques such as nrr to reduce validation well nir pretty much eliminates lab variation because you're going to get if you rescan things you get essentially the same number but that that's not that important nir if it's done correctly has to me it makes the cost the analytical cost low enough where you can afford to take enough samples so i think nir will will help reduce sampling variation because again you you can afford to take duplicate samples and you can sample more frequently so nir i don't think um it it's going to it's helpful because it's cheaper than wet chemistry analysis and you know people don't like to spend a lot of money and if nir cuts analytical cost in half they're more they made instead of sampling that may encourage people to take twice as many samples and that helps reduce sampling variation and adjust diet formulation there's you know there's some uh suppliers will have actually on-farm nir's where you know you buy an nr you could sample feeds every day at this point i'm a little leery on that um because you know these major labs will have calibration equations based on a million samples maybe their their calibrations are very very good but on a single farm you know your calibrations may not be so good so i'm i'm a little leery right now of on on-farm nir i'm very much in favor of of commercial lab nirs because again it should encourage more sampling more analyses if you find more questions please type your question from the q1a uh doctor lewis i have not uh another question about today you use the uh glue protein safety factor 1.8 i just want to know how you got this number 1.8 based on the study or based on your own experience so that was it's based both on on statistics on distributions and and some simulations we we you know we've we collected a lot of variations a lot of standard deviation then we'd sit and simulate diets just simulate diets and diets and if anyone's interested i have a uh excel program that'll do that and our goal was to say we wanted to have the same risk as if um if you fed very consistent you know corn if you had a diet with corn uh corn grain soybean minerals we didn't really include uh very good consistent corn silage but in other words we we said how variable is the best we can do using consistent feeds and then we'd simulate with these variable diets how we had to adjust it to get the same risk of a deficiency so it's based on on mostly stimulation and 1.8 was where a 1.8 times inclusion rate came up where the risk was similar to this this consistent diet the risk of a deficiency was the same as a consistent diet thank you very much yeah we scale more time for questions please type your question through the q1a okay there uh question for dr martin naturality excellent presentation your state that calls can handle a lot of variation is that because they'll draw on the body observers or change the efficiency with which they utilize the nutrients okay well i think it's a few things here one is you know we as nutrition researchers we do everything basically day to day we measure milk every day we measure intake every day etc but you know if if i change the diet on if i feed a diet on monday and i make a change on tuesday you know at least a third of the diet i fed on monday is still inside the cow so the just the capacity of the the gi tract in the room and buffers some of this day-to-day change and then the other thing what you you mentioned about reserves is i really think it's mostly of reserve uh thing and in a in this study that i i cited here we measured a lot more than what i talked about and on those days when i'm gonna just back up here so you just see on these days when we fed very high forage ndf the cow ate less and then what we on some measurements we did that it indicated that that cow when she ate less increased body mobilization so there was data that she increased uh energy mobilization fat so i think a lot and and we know on electrolytes for example they can do that um so i i think it's in the short term where we see this day-to-day short-term variation i'm guessing most of the the buffering is in reserves not in a change in in efficiency long term it might be inefficient for example with minerals many many days i feed low a calcium deficient diet she's going to operate upregulate absorption increase what i'm calling efficiency so i think over longer times the cow can moderate variation through through efficiency changes or at least moderated somewhat but day to day i think a lot of it is reserves and again a lot of it some of it is is the fact that you know there's care there the rumen and the intestines still have about at least a third of the diet she ate yesterday insider so another question from the yanxing new great presentation as in the presentation discount fact for the crude protein and the macrominerals is 1.8 so how about other nutrients the the other the reason we we're limiting it right now to protein and macro minerals and again i want to be clear this this 1.8 was an example you know it varies depending on inclusion right but um the reason we we use that for crude protein and and macro minerals again is the main risk with protein and macro minerals is deficiency with within reason in other words over feeding protein has an economic cost they their diet's more expensive but it doesn't hurt the cow and the same with most macro minerals it doesn't with within reason over feeding does not hurt the cow but a deficiency will but for these other nutrients there's a much narrower range between good and bad and let's take starch for example if if you want to feed a 26 starch diet and you have a highly variable feed you do my discount the average diet is going to be higher than 26 starch and that increases the risk for acidosis it reduces the risk of being energy deficient but it increases the risk of acidosis and so for other nutrients that the nutritionist really has to decide what risk is he want to take does he want to reduce the risk of deficiency then you would apply that discount but if on the other hand you say well i want to make sure i don't over feed this nutrient then you do just the opposite and if you're going to say well i want to keep the risk of over and under feeding the same you just you don't discount anything so these other nutrients it's up to the nutritionist to say which is worse a deficiency or an excess but for crude protein and macrominerals it's their deficiency is almost always worse than than overfeeding that's that's why we limit it to just those two nutrients thank you there any more question dr wales do you mind asking my last question so because uh you know our colleague and the students are made more against a nutritionist so the best your your experience with the handling the variation in the derick or nutrition do you have any suggestion for the motor gas trigger animal nutritionist how to handle the variation it's be very similar they have i'm going to say no the the non-room and nutritionists have it easier than us one is their feeds generally are less variable than our they don't have to feed forages and corn and soybean meal are pretty consistent and i think they also have they they can do much larger studies than we can so i think they have better variation or estimates and variation in in requirements and and so they can apply the same principles but there are the effects the diets are going to vary in in my opinion their diets will vary much less than ruminant diets and the animals are going to vary much less than than that so this all this i think would be less important for non it's still i don't want to say it's not important but it's going to be less important for those species than than our species our ruminants yeah thank you the question from dr marcus cadeno some people have talked about the animal level management should this be encouraged based on the fact that the animals can handle the variations the animal skin you know we've done studies and and there's been a few more people looking at this since we published this but we still need more data but the only nutrients we've looked at uh on the effect of variation is forage ndf this experiment experiment here uh in in journal dairy science we did a study with fatty acids and there's we've done a few studies with crude protein and i want in this study here the average which you know this is huge variation but the average diet was formulated to be 25 forage ndf which is quite high and we did that on purpose because i was really worried about sick cows and they handled this fine but you know in the real world you might be down here uh closer to the edge and then if you have that same variation you you might have a lot of sick cows so don't i don't want you to over interpret that variation doesn't affect cows the uh the stuff we've studied so far has not had huge effects but again we overfed 4g ndf in here so animals can handle more variation than we think but we need more more data with more nutrients under more situations before we say they can really really handle variation so um i still think animal these as we go we're getting better we have technology now where we can measure things on individual animals in groups you know the roman boluses the roman rumination monitor all this stuff so that will help on this but i i want to caution people to don't leave here thinking that cows variation do not does not affect cows for some nutrients it doesn't for other ones it it might and the experiment if i had if i had not retired the experiment i would have tried to get through animal care is doing starch and and starch you know our diets are very often 28 starch and i would have loved to do that i'm thinking we would have got sick cows because on these days when starch was 32 percent i think we'd have gotten sick cows and it would have been bad so if you're you're if you're young think about doing some of these studies so we get better data so i know i didn't really answer your question too well but um variation might affect cows more than this this one study shows yeah thanks for the question everyone asked the question and i think because the time interest so please join me thanks for dr phillips for the excellent presentation so i have learned a lot and i'm sure that our audience has learned a lot from your presentation yeah very much appreciate thank you and i will i'll i'll make these slides available i can go ahead and say i think i've already sent them to you feel free if someone wants them to distribute them i'd be no no problem on distributing yeah very much appreciate that and thank you again for the invitation good day everyone have a good day dr liz you too thank you very much goodbye