CALIFORNIA DEMO POSILAC RESPONSE ANALYSIS

MATHEMATICS OF ESTIMATING THE POSILAC RESPONSE

A very common method of estimating response is to use DHIA test day information. Cows are divided into two treatment groups, the POSILAC supplemented group and the control group, which are recorded as different strings in the DHIA data system. It is then a simple matter of looking at the difference between the two string averages to determine estimated response. Unfortunately, simple is not always correct, and in this case can be fraught with sources of error, which will negatively affect the response estimate downward from true respons

There are two common problems that arise using DHIA information that compares test day information on a calendar basis. One is an extended enrollment period (cows not all enrolled in POSILAC use on the same day). The other is differential pre-supplementation production levels.

An extended enrollment period introduces several sources of error into the estimation procedure, due primarily to the fact that on a single day, cows in both treatment groups are in different stages of lactation. Test day milk weights that contribute to a point on the graph come from cows that have not yet started supplementation with POSILAC, even though they have been designated a treated cow. Furthermore, early in the enrollment period, cows are through various stages of ramp up, which biases response downward. If all cows in the demo could be enrolled in one injection cycle this would not be an issue. However, in reality, most of the time, herd size prohibits this. Comparing cows using test number allows us to compare cows and treatments at similar stages of lactation. That is all cows are compared on their first test day, second test day and so on.

The matter of differential production levels prior to treatment application also introduces bias. Cows are sorted for demo group assignment on the most recent test day milk weight before entering the demo. Cows can have as many as 3 test day milk weights prior to entering a demo depending on the DHIA test interval and when enrollment began in the 57 to 70 day POSILAC start window. When an animal enters the demo they bring all their previous milk weights with them. This means that when evaluating differences in production levels by test numbers varying proportions of animals are receiving POSILAC in the supplemented group in test number 2 and often test number 3. In the POSILAC group we do know with a high degree of certainty that no animals are receiving POSILAC on the first test and all animals in the demo are receiving POSILAC on the 4t test. Thus test number 1 is used as a starting reference point since none of the animals in the demo are receiving POSILAC.

The easiest way to correct for pretreatment differences in production is to express all test day production as a deviation from the pretreatment level of milk production. As explained above, the only pre-treatment level of production known for all cows in a DHIA record system is Test Day 1 (TD 1) milk production. That is the case with the Demo herds summarized in this booklet. The following three graphs illustrate how these transformations are made. The graphs are from Demo A in this project booklet.

Graph 1: Average Milk Production By Calendar Day

Note:Numbers in red are how many Control animals were in milk on thetest date.Numbers in blue are how many animals were on POSILAC (1stnumber) and how many were in milk (2ndnumber) on the test date.

GRAPH 1: This graph shows the average production for treatment groups based on the date milk weights were measured. In this herd, cows were enrolled over a 6-week period starting on 6/16 and ending on 7/27. The earliest calving date was 3/8 and the latest was 6/1. This means that some cows were as much as 85 days in difference in their stage of lactation, a considerable length in time on any specific date. There were 14 control cows and 23 POSILAC supplemented cows that calved early enough to have a DHIA test on May 3. By the 6/7 test day, all enrolled cows had calved. But for 37 cows, 6/7 was their second test day, and for the remaining 70 it was their first test day.

Another issue evident in Graph 1. is that apparently control cows started out at a higher production level than supplemented cows.

If an average estimate of response across all dates was made from this data, the response appears to be 4 lbs.

GRAPH 2: Average Milk Production by Test Number

Note:Numbers in red are how many Control animals were in milk on thetest day.Numbers in blue are how many animals were on POSILAC (1number) and how many were in milk (2ndnumber) on the test day.

GRAPH 2: Data used to generate Graph 2, is exactly the same as Graph 1. No adjustments have been made to any milk weights; all milk weights have been used. Instead of summarizing averages by date of test, the data is summarized based on which test day it is for a cow. All first test days are averaged, regardless of the date for the test. In this graph all 55 supplemented cows have a test day 1 through 5 as do all 52 control cows. Because of the extended enrollment period, only 11 supplemented cows, and 9 control cows have a test day 6. The most apparent result of this realignment of data is in the shape of the curves. They now look like traditional lactation curves. This is because the data is now summarized on cows that are in approximately the same stage of lactation regardless of when enrollment occurred. Ramp up is occurring at only one point on the graph (TD 2).

This operation of realigning data to remove bias due to ramp up and differences in days in milk has removed 1 lb of bias, and our updated estimate of response now appears to be 5 lbs.

The only correction that yet needs to be made is for bias introduced because animals are at differing production prior to POSILAC supplementation. Control cows started at 5 lbs more milk (as measured at TD 1) than POSILAC supplemented cows. Thus, at any point in time along the lactation curves, estimating response by subtracting Control group production from POSILAC supplemented group production will yield a 5 lb underestimate of response. This can be easily over come by expressing production as a deviation from the last known production point prior to any cows being supplemented with POSILAC, which in this example is TD 1.

Graph 3: Milk Production Response Estimation

GRAPH 3: This graph is identical to Graph 2, with the exception that each Test Day weight is expressed as a deviation from the first test day weight using the procedure described above. In this way both groups now begin at the same point; when none of the animals in the demo are receiving POSILAC. Now the estimate of response for an individual test day can be expressed as a simple difference between the POSILAC supplemented group and the Control group on any particular test day. The weighted average of all test day differences (TD 1 and TD 2 are ignored because none or only a portion of some animals are enrolled in the POSILAC treatment group at these times) is now 10 lbs.

Source: Monsanto