This blog is in response to several requests from the clients...past and present...to start a discussion of Estimated Progeny Differences (EPDs). (We've noted that some authors call it Expected Progeny Differences. Either way, we are talking about the same concepts.)
I'd like to start off with a number of questions that Randy Bass of Moonlight Alpacas asked, as well as the responses I gave in an e-mail. Later, I hope to make some "how-to" videos using some cool technology I just discovered.
Meanwhile, if you have responses to this blog...either some new questions or comments, or something related to this particular posting, then please feel free to join in. This is the first real blog I've ever had, so, if there are some technical difficulties, please bear with me, and I'm sure we'll make this into a helpful and fun resource for all of us as we try to improve our herds.
Here goes...
Doug
I have another questions to add, you may have gone over.
When looking at the EPD's for consideration of breeding a
pair, is it the sum of the two. Does one animal maybe more
dominate have a bigger influence. Do we look at the accuracy to
determine.
DOUG: Great question! In a perfect world, the accuracy on an EPD
for an alpaca (let's call her "Alpaca A") would be 100% (1.0). If
it were, then the interpretation would be: Whenever Alpaca A has an
offspring, the measurement (be it Average Fiber Diameter, Standard
Deviation of Fiber Diameter, etc.) of the offspring is ***ALWAYS***
"X" amount less (or more) than the national average (NA) for the
national herd. For example, if Alpaca A's EPD for Average Fiber
Diameter (AFD) is -0.1 and the accuracy on her AFD EPD is 1.0, then
***GUARANTEED***, the offspring ***WILL*** have an AFD that is 0.1
microns less than the national average...NO MATTER WHO SHE IS BRED
TO.
So, already you can see the answer to some of your questions. The
higher the accuracy, the more the confidence, but it's not perfect,
because the accuracy is never 1.0. You can breed a given male and
female repeatedly, and the AFD of the offspring will be different.
That's because the AFD of the offspring is not 100% heritable (there
is variation from year to year, even with the same mating, due to
other factors, both known and unknown).
But, more to the point of what you are asking, yes, the higher the
accuracy, the higher our confidence that we can predict the quality
of the offspring. But, the accuracy is a mixed bag. An accuracy
can be low not only because the alpaca does not dominate perfectly,
but also because of a lack of data...and, this is important...the
statistics, as presented on the ARI site for Alpaca A, don't
differentiate as to the reason why the accuracy is low (or high).
The reason ARI doesn't differentiate is because, by the nature of
the analysis, they don't know, for an individual alpaca, how much of
the accuracy is due to dominance (OF THAT ALPACA), and how much of
it is due to a lack of data. (ARI does have statistics on how
heritable a characteristic is IN GENERAL, however.)
Here is another example: Suppose Alpaca A has PLENTY of offspring,
and all her offspring have had multiple fleece samples sent in to
Yocum-McColl, and that data has gone into the EPD calculation. But,
suppose Alpaca A was bred to a different male every year. What if
the offspring's AFD was ALWAYS similar to what the male had, every
time, and never the same as her? If that was the case, then the
accuracy for Alpaca A's EPDs would be low, NOT because the amount of
data is low, but because the measurements of her offspring are all
over the map. She does not dominate. Likewise, suppose she bred to
the same male every year, but the measurements were still all over
the map. Again, her accuracy would be low, and she does not pass on
her attributes to her offspring with any kind of consistency. But,
suppose her offspring have very consistent measurements...as
reliable as clockwork, whether she is bred to the same male every
year or not. Then she will have a high accuracy, because she
dominates for that characteristic, but it is unlikely that it would
be 100% (1.0).
GENERALLY SPEAKING, the higher the accuracy, the more that alpaca
will dominate for a particular measurement. However, that's not
always true, because, remember, just because a male she breeds to
has a low accuracy, that low accuracy may be because of a lack of
data, not because he fails to dominate.
How about the animals that don't have EPDs, and you find
relatives that do can we use the sum of the dam and sire? I
know the accuracy is a factor however if you have that animal
that you believe to have potential how do you come to a
conclusion to use the animal and what is the match.
DOUG: Another great question! The short answer is, if your alpaca
doesn't have an accurate EPD (presumably, at this point in time,
because of a lack of data), then there is no way to squeeze any
additional information, or to intuit an EPD, for the offspring.
Why? Because it is precisely the "squeezing of information" or
"intuiting an EPD" that ARI's EPD calculation does. In other words,
if an EPD could be "forcasted" based on relatives, then the ARI
system would PROVIDE you a forcasted EPD, with a calculated
accuracy. The fact that there is no EPD for Alpaca A means, despite
the EPDs of all its relatives, a reasonably reliable EPD for Alpaca
A cannot be derived.
The reason I can say this with confidence is that the very procedure
you are trying to use ***IS*** the procedure that ARI uses. They
take into account the parents, grandparents, aunts, uncles,
children, grandchildren, nieces and nephews to derive the EPD. If,
after the ARI system considers all of this, a reasonably accurate
EPD cannot be derived, then there is no other way to determine the
EPD with reasonable accuracy, using the math models in the system.
If you were looking at breeding to a sire off your ranch and
paying for a breeding would you ask for the EPDs.
DOUG: YOU BET!! Now, I would hasten to say, there are non-fleece
characteristics that come into play, as well as other fleece
characteristics, that are also important. Non-fleece
characteristics include, obviously, conformation issues, as well as
milking ability, color of fleece, etc. Other fleece characteristics
include luster, and, I would also say, density. Some would say that
the EPDs, which, when used in the right combination, can tell you
what kind of density to expect, and that is true. But, there is no
substitute for measuring density directly (through a skin biopsy),
and an EPD using a skin biopsy is not in the system today.
So, yes, I would consider EPDs, but it's not the whole picture.
We need more ranches to release their EPDs.
DOUG: Amen! Preach it, brother! Talk it up to other breeders.
They may have reasons why they are not releasing them, which is
fine. But, if it is a result of unawareness of the system, then we,
each of us, need to talk to them and get them to make their EPDs
public.
That said, if someone has not made their EPDs public, you can still
ask them to privately look at their male using their own account and
have them produce a PDF file of their EPDs. If you want more info
on how to do that, let me know.
I have answered some of the questions and will work on the
others.
6. Yes the accuracy is a percent even know they print it in
hundredths and the column is in percents.
DOUG: Yes, it is a percent, but a percent of what? That is a large
topic that I will be discussing at the next meeting.
8. The national average is the actual value, and therefore
zero would mean that the would have to recalculate all animals
every time the NA changed to reflect how a particular animals
standings are. As the NA changes up or down the individual
animal remains the same, however the accuracy of that animal
could change.
DOUG: I'm not sure I understand what you are saying. Yes, a given
alpaca's EPD could change from year to year, as more data are
entered into the system. However, you can't just change the
National Average to adjust for that change, because one alpaca's EPD
may have gone up by 0.1 microns, but another one's may have gone up
0.2 microns. One change to the national average can't account for
the variation of all the alpacas that have EPDs.
ARI DOES recalculate the EPDs every year for every alpaca. The
individual alpacas' EPDs don't remain exactly the same year to year,
as a rule.
Again, I might not be understanding what you are saying, and if I
don't, I apologize.
Has anyone else found out more information about this?
Luster is a characteristic than can be controlled to some
extent by diet and environment so to rate this would not be
rating the animal so much as the management of the ranch.
DOUG: Agreed. Luster can be controlled ***to some extent*** by
diet. So can average fiber diameter, the standard deviation of
average fiber diameter, and, perhaps all of the attributes. But,
there is still a considerable extent by which luster (and the other
attributes) is heritable. Sure, nothing is 100% due to heritability
(except perhaps fleece color, eye color, and things like that), but,
with our breeding decisions, we are trying to control the
heritability part of the equation.
I would say that luster is reasonably heritable, so therefore I
believe it is a worthy goal to try to develop a luster EPD. Anyone
else have any thoughts?
That is what I've come up with so far I'm very interested in
your answers for all the questions. If I get more I'll let you
know.
This is great that you are doing this I only wish I was 300
miles closer.
DOUG: Well....if you'd like, we could get you in on a speakerphone
for the meeting. Sometimes that's a little difficult, not being in
person. But, let us know, and we'll see what we can do!
Thank you for keeping me in the loop.
DOUG: And thank YOU for your participation!
