OFR21: Intro to On-Farm Research with Nathan Mueller

Nathan Mueller Author

10/26/2021 Added

32 Plays

Description

In this video Nathan Mueller discusses the basics of conducting your own on farm research project and how it can add value to your operation.

Searchable Transcript

Search:

[00:00:00.196](uplifting music)
[00:00:09.960]I'm a cropping systems extension educator,
[00:00:12.290]just like many of your local hosting educators.
[00:00:15.340]And for those of you online,
[00:00:17.290]we're spread across the state,
[00:00:19.010]and I'm located in Jefferson and Gage County.
[00:00:21.520]So I'm gonna go through
[00:00:23.170]how to read the On-Farm Research results table.
[00:00:27.720]And so that way we kind of are all on the same page
[00:00:29.760]moving forward for the day.
[00:00:32.230]So basic understanding.
[00:00:33.580]So this is just an example out of the book from page 19,
[00:00:37.500]but just wanted to walk through
[00:00:38.990]what some of these numbers mean.
[00:00:40.180]And I think probably the main point
[00:00:41.710]I wanna drive home on this, what is a P-value?
[00:00:45.560]any of you that watch Parks and Rec,
[00:00:48.160]you'll recognize this character,
[00:00:49.670]but a P-value stands for probability value.
[00:00:53.500]And essentially, it's a test that we gotta conduct
[00:00:56.840]that's called an analysis of variance,
[00:00:59.020]and the test is called an F-test,
[00:01:00.670]and then we get this P-value.
[00:01:02.730]What we're testing are all the means
[00:01:05.510]or all the yield results
[00:01:06.760]of these four different seeding rates,
[00:01:08.710]are they equal or are they not?
[00:01:11.830]So if you look at stand count,
[00:01:13.460]yep, those don't look the same.
[00:01:15.930]We go over to yield and we see 87, 88, 87, 86.
[00:01:20.880]Are those equal or not?
[00:01:22.130]So we conduct that test.
[00:01:23.660]And if the P-value is less than 0.1 or 10%,
[00:01:28.350]then we say they're equal,
[00:01:31.240]or they're not, they're equal.
[00:01:33.230]If the P-value is less than 0.1 or 10%, they're not equal.
[00:01:39.350]So if we look at the P-value for yield, it's 0.38.
[00:01:43.890]It's larger than 0.1,
[00:01:46.070]so those means are equal or there's no treatment difference.
[00:01:50.320]If we go over to stand count,
[00:01:51.880]since we had a seeding rate steady,
[00:01:53.560]we would hope that there's a difference.
[00:01:56.180]And the P-value there is less than 0.0001.
[00:02:00.510]So significantly less than 0.1, our cutoff level, or 10%.
[00:02:05.420]Now, that 10% cutoff level for determining
[00:02:08.220]if something's significant or not
[00:02:09.690]goes back to what Joe talked about,
[00:02:11.250]that random error or noise that we can have in the data.
[00:02:15.320]And we wanna make sure that we're drawing
[00:02:17.030]the correct conclusions, that the yield difference
[00:02:19.390]or the treatment difference we're seeing
[00:02:21.010]is due to the treatment,
[00:02:21.960]not that random error or noise in the background.
[00:02:24.620]So the 10% level is arbitrary.
[00:02:27.970]And we'll talk about another slide,
[00:02:30.130]an example of where somebody used a different cutoff level.
[00:02:33.240]So next slide.
[00:02:36.520]So just to reiterate that, in your book,
[00:02:39.860]and we'll give you access too, to the online one as well.
[00:02:43.310]Laura will be sharing that, but on page eight,
[00:02:45.810]there's a summary of kind of what we're covering.
[00:02:47.580]So if you ever need a refresher, read through page eight,
[00:02:51.390]but just a reminder,
[00:02:52.720]we're using a 0.1 or a 10% cutoff level.
[00:02:56.070]So again, if the value's 0.11,
[00:03:00.040]approaching significance here,
[00:03:02.962]the third bullet point,
[00:03:04.700]technically, you're gonna see all the letters,
[00:03:07.050]no difference.
[00:03:07.883]It's gonna be the letter A for all the treatment different.
[00:03:10.040]So we're approaching significance,
[00:03:11.870]but it's not quite there.
[00:03:13.820]At 0.25, we're gonna say, there's no treatment difference.
[00:03:17.060]You're gonna see the letters are all the same.
[00:03:19.480]We're assuming at that point,
[00:03:21.280]that that difference there is likely just random error
[00:03:24.580]or noise that Joe talked about in his presentation,
[00:03:28.110]that can come from the combine artifacts
[00:03:30.890]or the combine or field variability.
[00:03:33.370]Next slide.
[00:03:36.490]So that was the first step was,
[00:03:39.280]are the means or are the treatments equal?
[00:03:42.210]And then we needed to know the next step,
[00:03:44.100]where are they different?
[00:03:45.280]What treatment is different than the other?
[00:03:48.400]So that's the next test.
[00:03:50.200]And that's called a mean separation test.
[00:03:52.480]And we use Tukey's, which does protect against error rate
[00:03:56.960]when we make lots of comparisons.
[00:03:58.650]So if we have six different treatments,
[00:04:00.470]or four in this case,
[00:04:02.150]the chances us making an incorrect conclusion
[00:04:04.690]go up with each individual pairwise comparison.
[00:04:07.950]So we are controlling for that,
[00:04:10.890]so that way, again,
[00:04:11.760]we don't attribute a treatment difference right away,
[00:04:14.240]just that may not be true.
[00:04:16.830]So if you look at the letters
[00:04:17.990]that you're gonna see in all the tables,
[00:04:19.390]it's just like Sesame Street.
[00:04:21.480]If the letters are the same, the treatments are the same.
[00:04:24.110]There's no difference.
[00:04:25.950]If the letters are different,
[00:04:27.090]then we're seeing a difference according to the statistics.
[00:04:29.620]So Laura's highlighting there, letter D, C, B, A.
[00:04:33.180]So actually, we are able to separate out
[00:04:35.410]with the statistical counts that we got that,
[00:04:39.270]yep, we hit those.
[00:04:40.400]There's changes in the seeding rate across the field.
[00:04:43.320]If we go over to yield,
[00:04:44.980]you can see with this mean separation test,
[00:04:47.270]we got all letters A.
[00:04:48.420]So that one or two bushel numeric yield difference
[00:04:51.330]you're seeing there,
[00:04:52.300]we're not confident that that two bushels
[00:04:54.330]is actually due to the seeding rate.
[00:04:56.850]And then we move over to marginal net return.
[00:04:59.240]And we'll talk about how we calculate this,
[00:05:01.490]but we have A, A, AB and B.
[00:05:04.220]So what's this AB?
[00:05:06.310]It's like A and it's like B.
[00:05:09.320]We weren't quite able to separate out
[00:05:11.380]the 140,000 seeding rate from the 110 or the 170.
[00:05:15.880]We're not confident that that change in seeding rate
[00:05:20.490]is gonna drive a marginal net return difference.
[00:05:23.720]But we are confident that the 80,000 seeding rate profit us
[00:05:27.940]or had a higher and marginal net return
[00:05:29.910]than the 170,000 seeding rate
[00:05:32.230]because the difference in the letter A and B.
[00:05:34.700]If any of you have followed FIRST Trials,
[00:05:36.280]it's a third-party entity in Eastern Nebraska
[00:05:38.970]that does variety testing.
[00:05:40.090]This is from their plot this year in Beatrice.
[00:05:42.530]The plot average is 209.
[00:05:44.820]Sometimes you'll see a value called LSD
[00:05:47.050]and that's least significant difference.
[00:05:49.160]It's how many bushels there needs to be a difference
[00:05:51.980]between two hybrids, or in this case,
[00:05:53.770]two different seeding rates
[00:05:55.530]for it to say that it's due to the hybrid,
[00:05:57.240]it's due to the seeding rate.
[00:05:58.820]So they report both, a 0.1 net 10% cutoff level.
[00:06:03.200]And then they also report a 25% or a 0.25 cutoff level.
[00:06:07.730]And of course, you only need a 3.8 bushel difference
[00:06:12.010]at that lower confidence level.
[00:06:14.760]However, you also have a 25% chance of being wrong
[00:06:18.240]in drawing a conclusion that that four bushels
[00:06:20.550]was due to the hybrid in their plot,
[00:06:22.700]where at the 7.3 bushels,
[00:06:24.340]you only have a 10% chance, essentially, of being wrong,
[00:06:27.960]that there actually wasn't a,
[00:06:30.210]that you're saying there was a hybrid difference.
[00:06:32.100]So again, it's about being confident in the results,
[00:06:34.910]and that's what we're using statistics for
[00:06:37.220]is to be confidence in the conclusions that we're drawing.
[00:06:40.620]So next slide.
[00:06:43.710]So all of the yield data that you're seeing,
[00:06:46.420]and then that goes into the marginal net return,
[00:06:48.610]we have corrected for moisture across crops.
[00:06:50.870]So for base-level moisture that we're correcting for,
[00:06:53.660]for soybeans here is 13%.
[00:06:55.440]So if for a treatment, the percent was 12.5%
[00:06:59.480]in one treatment or one plot and the next one was 13.5,
[00:07:03.940]we're adjusting the total weight in the bushels
[00:07:06.320]to standardize everything at a common moisture percent
[00:07:09.180]across all of these studies.
[00:07:12.000]So next slide.
[00:07:14.750]So the marginal net return calculation that we use
[00:07:18.170]that you'll see in all the tables
[00:07:19.530]is gross income minus the treatment cost.
[00:07:23.570]And so let's run through, for the 80,000 seeding rate,
[00:07:27.510]what the marginal net return was.
[00:07:30.360]So it would be our yield difference
[00:07:32.120]or our yield measured for that treatment, which was 87.268.
[00:07:36.230]We do, in the tables,
[00:07:39.940]average that out to a full bushel difference.
[00:07:42.910]So if you're calculating this at home by calculator,
[00:07:46.830]you might see a little bit of a difference
[00:07:48.490]by a dollar or two, but we use that full number that we get.
[00:07:52.670]So we take 87 bushel times $9.50,
[00:07:56.520]which is a standard commodity price
[00:07:58.190]that we're using this year in 2020,
[00:08:00.320]minus our treatment cost,
[00:08:01.720]which was 80,000, divided by our 140,000,
[00:08:05.900]which is our unit for soybeans,
[00:08:07.250]times the unit price, which was $62.30 a unit.
[00:08:11.820]So we just do the math on that.
[00:08:13.270]It's $829 minus $35
[00:08:16.910]and we get our $793 marginal net return
[00:08:20.470]that you see in the table.
[00:08:22.030]So essentially, you can verify these as you go through,
[00:08:24.710]if you happen to see a little bit of a blip
[00:08:26.670]or a number that isn't doesn't quite make sense,
[00:08:28.140]sure, let us know.
[00:08:29.960]Sometimes we don't catch all of the errors
[00:08:32.600]when we're running math.
[00:08:33.500]So next slide, we'll talk about
[00:08:35.540]how the numbers we're using for this.
[00:08:38.280]So for the marginal net return calculations,
[00:08:41.010]again, on page nine in the book,
[00:08:43.580]this information is shared there too,
[00:08:45.420]but the average commodity prices we use
[00:08:47.340]for the 2020 studies was 3.51 bushels per acre,
[00:08:51.987]$9.50 per bushel for soybeans,
[00:08:55.457]$6.01 roughly per bushel for cereal rye,
[00:08:58.700]and pinto beans, $24 per hundred weight.
[00:09:03.440]Now, the costs that we use on the studies
[00:09:07.050]come from several different sources.
[00:09:08.470]It could be the producer-reported cost of that treatment,
[00:09:12.190]what they bought those soybeans for.
[00:09:14.550]And when it comes to application,
[00:09:16.610]if we don't have the applicator's cost,
[00:09:18.970]we will use the UNL custom rates to substitute for that.
[00:09:22.960]We will use the 2020 commodity prices
[00:09:25.710]for the marketing year, not necessarily what the farmers
[00:09:29.080]actually sold their crop for, just to standardize that.
[00:09:32.710]And then of course, you can use your own data.
[00:09:34.710]So if you're looking at the results of one of these studies
[00:09:37.210]and you wanna input some of your own costs or your own,
[00:09:40.320]what you got for your commodity
[00:09:41.800]or use your application cost instead of the custom rates,
[00:09:44.890]you can sure do that and kind of rerun some of the math
[00:09:47.780]on your own for each of these studies.
[00:09:50.540]So with that, that kind of wraps up
[00:09:52.390]about understanding the results table
[00:09:55.150]in the On-Farm Research book.
[00:09:56.610]Thanks for your time.
[00:09:58.272](uplifting music)

The screen size you are trying to search captions on is too small!

You can always jump over to MediaHub and check it out there.

Comments

0 Comments

OFR21: Intro to On-Farm Research with Nathan Mueller

Description

Searchable Transcript

Comments icon comment

Related Channels

Comments