A FIELD GUIDE for ON-FARM - RESEARCH EXPERIMENTS
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS The Cooperative Extension Program North Carolina A&T State University
GLOSSARY for On-Farm Research Experiments WHAT WILL I GET FROM ON-FARM RESEARCH?
Data: Data is information Plot: A plot is an area of specific Farming is a challenging
that you collect in a research length and width (such as 4 rows occupation. If the weather
experiment. The data can be or 2 plant beds that are 100 feet isn’t giving us a headache,
quantitative (numerical, e.g., long) where a treatment takes other problems or puzzling
pounds of yield) or qualitative place. questions always crop up.
(evaluative, e.g., superior, good,
Sometimes solutions or
fair, poor). Yield data from a Guard Rows: The outer rows,
answers are provided to us by
crop is commonly collected called guard or buffer rows,
INSIDE to compare different varieties protect the inner rows from the
friends, neighboring farmers,
county agents, or from other
or practices. influence of different treatments
in neighboring plots. Data is agricultural professionals.
What are the
drawbacks to Treatment: A treatment is a usually collected from inner But other times answers are
on-farm research? ............2 production practice that you rows in the plot. not readily available. The
want to investigate. Usually, an question or problem we have
What question on-farm experiment compares Randomization: In a side-by-side may be peculiar only to our
do you want the two practices: the standard comparison, randomization
research to answer? .........2 own situation or specific to
practice and a new practice. assigns Treatment 1 and Treat- our own site. No one may
Field Variability ................. 3 For example, you may want ment 2 to their field position
have any answers.
to compare: 1) the yield from within a block (e.g. to the left or
In all probability, others
Replication ........................ 4 a standard variety of beans right of each other) by chance,
may have asked similar
with the yield of a new variety; such as by the flip of a coin.
Randomization ................. 4 2) full tillage vs. strip-tillage; questions. They, too, may be for himself? Can you expect with other farmers. When
or 3) starter fertilizer vs. Variability: Variability describes frustrated that answers are to see higher yields on your farmers do research on their
Plot Size and not available. For example,
no starter fertilizer. how conditions change in the farm, also? own farms, they are, in effect,
Data Collection .................5
experimental field. For example, university researchers recom- Here’s another example: taking responsibility for
Common Questions .........6-7 Block: A block is the basic there might be wet spots in the mend a new bean variety. Every year you apply an solving their own problems.
unit of an experiment. A block field or the topsoil might get If it yields poorly in one herbicide, for weed control, In actuality, farmers
Steps to a Successful contains two (or more) different deeper as we move down the particular county, growers before planting beans. You are always engaged in some
Research Project ...............8 treatments in a side-by-side slope. Since most fields are in that county will probably usually follow this application type of research on their
comparison. A block is one variable, we replicate and
On-Farm Research want some answers. Are the with a second herbicide farms. They try something
Helpful Hints .....................9 replication of the comparison randomize to minimize the
soils or the climate different application after weeds new (tillage practice, fertility
of treatments. variability within a block.
causing lower yields? Are come up later in the season. scheme, variety, etc.), observe
Statistics ............................10-11 Variability can influence the
Replication: A replication is response we see from the “conventional” soil and Herbicides are expensive the effects of that new thing
Example Statistical a side-by-side comparison that treatment (data), and therefore fertility management practices and you’ve often thought, (effects on yield or income),
Calculation ........................12-13 is identical to and takes place in the conclusions we draw in that particular county “What if I skip the pre-plant and draw conclusions from
the field at the same time as the from the experiment. inappropriate for the new herbicide to save money?” what they observe (the new
original side-by-side comparison. variety. How will you determine if this practice is or is not worthwhile).
Normally a research experiment Statistics: Statistics is a Say your neighbor is is an effective strategy that On-farm research can be
has four to six replications mathematic way of comparing promoting a “hot new bean will save you money, but still as simple as an observational
A Field Guide for On-Farm (equivalently, 4 to 6 blocks). the data collected in all the variety” with “higher yield.” give you the yield you expect? study. With a little more work,
Research Experiments The more replications that you replications. It allows us to
How does your neighbor know These examples present growers can increase the
have in your experiment, the draw “scientifically valid”
By Keith R. Baldwin, Ph.D. that it will produce a higher situations where growers may certainty that the conclusions
more confident you can be in conclusions from our
Horticulture Specialist, yield than the standard vari- want to do their own on-farm they draw from their observa-
the results. The replications comparisons.
Cooperative Extension should be as similar as possible ety? Did the seed company research. They seek answers tions are correct. Thus, they
Program at North Carolina with respect to field conditions representative say so? Was to their own questions, or, can reduce the risks associated
A&T State University and agronomic practices. this claim based on scientific when there are questions with making changes that
research? Or did your neigh- common to a community, have potential for increasing
Greensboro, North Carolina bor grow that variety to see they may work cooperatively farm profitability.
March 2004WHAT ARE THE DRAWBACKS TO ON-FARM RESEARCH? FIELD VARIABILITY
Make no mistake, on-farm in a single year. You may plot sizes helps to limit any In a very simple variety Field variability can be If we are merely curious
research requires planning discover that the old variety financial risk to capital and trial (Figure 1), we could plant present when: about how a new variety will
to get the information you or the “tried and true” practice labor that may accompany any half of a field to a new variety perform or if a new practice
• Erosion has moved the topsoil
seek. It may also require some is best. If a large area on the experimental “misfortune.” of beans (Treatment 1) and is better than an old one, a
down slope in the field
additional effort and time. farm has been dedicated to You (or someone you know) the other half of the field to side-by-side comparison may
• Wet spots are present in
You will need to remember the research, there may be must be able to do some sim- the standard variety of beans give us all the information we
the field
to perform research tasks on a substantial investment in ple math to make sense of the (Treatment 2). We would need. In this kind of observa-
• Soil moisture increases from
time. Unfortunately, many of research that merely validates information (data) you collect. harvest beans from Treatment 1 tional study, however, we
one end of the field to
those tasks will take place current practices. This is so that you can get a and weigh them. Then we cannot be certain that field
the other
during a very busy harvest The research may even scientifically reliable answer would harvest the beans from variability didn’t bias the
• The soil may vary from
period. This is usually the result in a total crop failure. to your research question Treatment 2 and weigh them. results that we observe. For an
sandy to loamy to clayey
case when you are collecting That great new variety your and have confidence that the Finally, we would dry the beans observational study, we should
throughout the field
yield information to answer neighbor told you about might answer you get is correct. because the treatments may always try to pick a field site
• Parts of the field get
your research question. Extra grow like gangbusters in a Fortunately, there are some have differing amounts of mois- that is as uniform as possible.
afternoon shade
work may be required at this year with normal rainfall. simple worksheets that can be ture. If the “dry weight” yield of If we want to reduce the
• Weed pressure is higher in
time. For example, harvested But it may fail miserably when used to do the mathematics the new variety is greater than chance that variability influ-
one part of the field
crops from each research plot conditions are drier. The stan- (statistics) that only require a the standard variety, then we ences results and to be more
• Bean-eating deer feed in
must be weighed separately. dard variety may perform rea- calculator. Or, for those with might conclude that the new certain in our conclusions,
the section of the field
The project itself can be sonably well in wet and dry computers, a very simple pro- variety is better. we need to have several
nearest the woods
somewhat risky. Recognize years. Often this explains gram — where you “fill in the How can we be sure that replications or blocks.
that your research may not the reason it is a “standard blanks” — can do the statistical any difference such as higher
provide any useful information variety.” The use of smaller calculations for you. yield in the Treatment 2
plot is because of the
variety or practice? The Figure 1.
WHAT QUESTION DO YOU WANT THE RESEARCH TO ANSWER? answer is that we can’t. Side-by-side observational experiment.
There may be other
Beginning, on-farm researchers should keep are a few more common questions that reasons that explain the
research projects simple and choose just one growers can ask to increase farm income: difference in yield. For
question to answer. Commonly, the question example, the new variety
will ask whether a new way of doing things is • Will my yield increase if I irrigate may have been planted
“better” in some way than the current way of my sweet corn? in better soil than the
doing things. The answer to that question will • Can I substitute a green manure cover standard variety, some-
either be “yes” or it will be “no.” For example, crop for commercial fertilizer without thing that often occurs
your research question might be “Will this new lowering yield? when the soil in a field
bean variety produce a higher yield than the • Can I eliminate a particular pesticide is variable. In Figure 1,
standard bean variety that I usually plant?” application and increase my bottom the Treatment 2 is in
The answer in this simple variety trial line per acre? better topsoil than the
will be either, “yes, it does” or “no, it doesn’t.” • If I plant my pumpkins no-till, will Treatment 1 plot.
Obviously, if the answer is yes, there is an my yields be reduced? We can naturally
opportunity to increase farm income by expect crops to grow
planting the new variety of beans. If the new Think about the research question as a better there. Treatment 1 Treatment 2
variety yields five bushels more per acre comparison. The examples above compare
than the standard variety, income could two practices: 1) irrigation vs. no irrigation; Slope
be increased by over $20 per acre! 2) a cover crop vs. commercial fertilizer;
Most farmers have some experience 3) pesticide vs. no pesticide; and 4) tillage vs. Moderately Highly
with variety trials like the one described above. no tillage. In each case the practices compared Productive Soil Productive Soil
There are many other yes/no questions that are called “treatments” and the on-farm research
can be investigated in on-farm research. Here will compare Treatment 1 with Treatment 2.
2 A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS 3REPLICATION PLOT SIZE AND DATA COLLECTION
On-farm research experiments are Treatment 1 and four more rows of Treatment 2 The size of the treatment through winter dormancy. can decide for yourself how
“replicated” to address the issue of field create a third block. We usually continue creating plot is determined by the Since the pesticide spray long the plots should be to
variability. To have confidence in our answer new blocks until we have a minimum of four size of the equipment might drift from the spray provide a representation of
that “yes the new beans are better,” or “no they blocks and, preferably, a total of six. you will be using to do the treatment to the leaf removal a field situation.
aren’t,” we normally repeat the experiment four Even with replication, in a replicated experi- fieldwork, in particular, treatment, at least one row of Place the harvested prod-
to six times! This doesn’t mean that we need to ment, we can expect the yields in six different harvesting equipment. For strawberries should be left as uct from each treatment in a
repeat the experiment for six years. We usually Treatment 1 plots to be different from one example, if you are using a guard row between the separate bag or container. In
conduct all six at the same time, in the same another because of field variability. The same is two-row equipment, the plots treatments. After all the yield a block with two treatments
field, and in the same year. Because weather can true in the Treatment 2 plots. When we have six will need to be at least four data is collected from all the there will a bag for each
affect crop performance, we often conduct the blocks, we may even see that Treatment 1 yields rows wide. In a four-row plot, plots, the guard rows can be treatment. In an experiment
experiment two years in a row. are greater than Treatment 2 yields in some the two-row harvester picks harvested. with six blocks you will have
Each repetition (or replicate) is called a block. blocks, but not in others. Replication helps us the two middle rows to get When working with bed- two bags in each block and
Taken together, four rows of a new variety of decide whether the treatments themselves caused yield data. The outer row on ded crops, such as strawber- the total number of bags will
beans (Treatment 1) and four rows of the standard the yield differences. It helps us to rule out field either side is called a guard ries, two beds commonly be 12 (2 treatments X 6 blocks
beans (Treatment 2) are a block. Four more rows of variability or some other random event as the row. The guard rows protect make up one plot. Harvest = 12 bags). Label the bags so
Treatment 1 and four more rows of Treatment 2 cause of any difference. the inner rows from the data is collected from the that you will remember
create a second block. And four more rows of influence of different treat- inside rows of each bed. which treatment and block
ments in neighboring plots, When hand-harvesting crops each one belongs to. Weigh
RANDOMIZATION and are not harvested when planted in single rows, data the bags and record the
collecting yield data. is sometimes collected from weights. In some cases, the
We also address field the upslope (left) side of each there are six blocks. Each block Guard rows are most one row. In this case, the bags of harvested product
variability by randomizing block, Treatment 2 will always is identical because it compares important in experiments harvested row usually has will be completely dried after
the treatments in each block. have better soil than Treatment 1. Treatment 1 with Treatment 2. where one treatment may guard rows (or guard beds) on harvest. As drying proceeds,
By mixing up the placement So we randomize the treatments However, the positioning of the affect the other. For example, both sides. moisture will be lost from
of the two treatments in each in each block so that we have an treatments in each block is not a strawberry grower may The length of the plot is the product and it will lose
block, we adjust for any advan- experimental blueprint like that identical. Sometimes, Treatment want to compare two prac- commonly the length of a weight. When dry, the bags
tage the left or right position shown in Figure 2. 1 is on the left side of the block, tices for managing gray mold row (row crops) or a bed (veg- will be weighed again to get
might enjoy. In Figure 1, the soil In on-farm research we try and sometimes it’s on the right in strawberries. Treatment 1 etable crops). However, you a “dry weight.”
gets better as we move down the to avoid any influence of field side. You can assign a side for might be a pesticide applica-
slope. If we place the blocks side variability by replicating and Treatment 1 in each block by tion and Treatment 2 might
by side down the slope, and if randomizing within each replica- tossing a coin: “heads, it’s left,” be the removal of dead
we always place Treatment 1 on tion. Notice in Figure 2 that and “tails, it’s right.” leaves that carry the mold
Figure 2. Replicated and randomized side-by-side experiment
1 2 2 1 2 1 1 2 1 2 2 1
Block 1 Block 2 Block 3 Block 4 Block 5 Block 6
Slope
Moderately Highly
Productive Soil Productive Soil
4 A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS 5COMMON QUESTIONS
Who might be able to have planned. Or, those What research methods can was better than another, and
help with the research? reports may provide informa- I use to insure that not because of some other
You would be surprised tion on where to get more the answers I get are true? factor. You do not want to
how many people would like information. Replicate your experi- have to wonder, following
to help you with your on- Other agricultural profes- ment by comparing the your experiment, if the high-
farm research project. In sionals will also be willing treatments in four to six sep- er yield was a result of sow-
many counties, Extension to help. State Departments arate blocks and randomize ing the test variety in a better
agents participate in on-farm of Agriculture, the Natural the placement of treatments field, or because of better
research projects. Contact Resources Conservation in each block. Replication weather the year the test
your local county agent for Service, other farmers, and and randomization will variety was planted. Nor do
help. Your agent has likely farm organization employees account for any differences you want to be left to wonder
had some training in on-farm have a wealth of information between treatments that may if the test variety might have
research and would be happy to share. Call on them to be caused by field or other produced better yield because
to participate in your project. help you with your project. sources of variability. fertilizer and pesticide inputs, nor from all Treatment 2 it was given more care and
County agents have The World Wide Web has Treat all the research and harvest dates should all plots. You will need the attention than the old variety.
resources that will be of created a treasure trove plots in the same way. be the same. If nothing is dif- results from the separate By all means, share infor-
value, too. They have access of information that can be The only difference between ferent except the treatments, treatment plots for a statisti- mation from your experiment
to libraries at universities, searched from a home com- any plots is the treatment. then any differences between cal analysis of the data. with other farmers. But be
and those universities may puter. Most public libraries Unless they define the treat- the treatments are likely to aware that the more blocks
have reports of research that have computers that are ments under study, planting be the result of the treat- Can I share my you have, the more confident
may be similar to what you connected to the Internet. date/depth/spacing, tillage, ments. conclusions with others? you can be that what you
This will help you get the By all means share observed is scientific fact.
best answer to your question. your information with other This is especially true if you
Sometimes we want a farmers. But, be sure to have randomized the treat-
research question to be share your “confidence” ments in each block.
answered in a certain way. in the information at the You can be even more confi-
For example, we may want a same time. You can be most dent in a recommendation
new variety that we “discov- confident if you have replicat- if you have conducted the
ered” to be a humdinger, and ed the experiment and ran- experiment in multiple years.
we may give it extra special domized the treatments
care relative to a common in each block. How important is picking
variety we’re comparing it to. For example, is it appro- out a site for my research?
Take care that any bias does priate to recommend a new Site variability is critical-
not influence the way you bean variety to your neighbor ly important in a research
manage the experiment. based on one side-by-side experiment. The more
study on your farm? If the uniform the site (soil, slope,
If I am collecting yield neighbor makes a decision moisture, etc.), the better.
information, what to plant his whole farm to Experimental sites usually
precautions should I this improved variety next vary in some way, and we
take at harvest? year, and he goes belly-up try to overcome this
Keep the harvested when the crop fails, how difficulty by replicating
product from each treatment will you feel? the experiment many
plot in every block in a sepa- When you make a times in separate blocks.
rate container and weigh recommendation from on- We randomize treatments
each treatment separately. farm research, you want to (left/right or right/left) in
Do not combine the product be confident that the higher each block to account for
from all Treatment 1 plots yield was because one variety variability in the block.
6 A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS 7STEPS TO A SUCCESSFUL RESEARCH PROJECT On-Farm Research HELPFUL HINTS
1. Put your research idea each replication, conditions the work will get done during
into a question. The question such as slope, soil texture, busy periods such as planting 1. Keep it simple. Start with a the results. Remember that results
usually turns out to be a exposure, drainage, etc. should and harvest. Will the crop or simple comparison, with a limited can vary from year to year. It is
comparison of two treatments, be as uniform as possible. the test be planted first? number of treatments. It is far important to have a long-term per-
and takes the form of, “Is Though it is desirable for Will extra equipment or labor better to have a simple trial that spective, and to wait until several
Treatment 1 better than conditions in all replications be needed? Who is going to yields clear results, than a years worth of data is collected
Treatment 2.” to be uniform, this is usually do what tasks? complex trial where results before drawing conclusions.
not possible. are confusing. Starting with a
2. Decide what data you 8. When you plant the simple research question is 5. Manage risk. Field experiments
will collect. In most cases 5. On the map, place the crop, remember to crucial. Rather than attacking a are a method for managing risk.
this will be the yield of each treatment plots side by randomize the treatments complex question all at once, If done well, they can give
treatment, but it also could side in each replication. in each block. most complex issues can be farmers a method for evaluating
be other information, such as Flip a coin to decide which broken down into a series of new production practices on a
the number of diseased plants treatment plot is on the left 9. Don’t trust your memory. simpler ones. limited scope, reducing the chance
in the plot or the number of side of each pair. If the coin Keep accurate records of of adopting a practice that doesn’t
insect pests on 10 randomly toss is heads, put Treatment 1 everything you do for the 2. Plan for busy periods. The most work. Take care that the size of the
chosen plants. on the left, and if it’s tails, crop and when you do it. crucial times for a field experiment experiment itself does not add too
put Treatment 2 on the left Record information about are also the most crucial time for heavily to the risk level. One rule
3. Describe what each side. If the blocks themselves rainfall and any unusual the crop. Planning the experiment of thumb is that a field experiment
treatment will be and a or the treatment plots within weather (storms, wind, or hail). during planting season and taking needs to be big enough that the
plot size that will be the blocks are separated from If the plan changes (e.g., data during harvest can both be farmer cares about the result, but
representative of a field each other by buffer zones, be fertilizer rates), record the very difficult, and require time not so big that they lose sleep
situation. Remember that sure to include these buffer changes and the reasons for when farmers least have it. Plan worrying about the money
all farming activities will be areas on your map. them. Sometimes facts that the experiment well in advance, wrapped up in it.
the same for both treatments. don’t seem very important and note how things are going to
Any differences you observe 6. With a tape measure lay at the time become very happen when the time gets short. 6. Get help. There are many
(such as increased yield) out the plots in the field. important in understanding agencies and individuals who
will be because of the Mark the corners of each why something did or didn’t 3. Watch for unexpected results. will be interested in the results of
difference in the treatment treatment carefully, so you go the way you expected. Changes in production practices your research. Don’t hesitate to
and not because of difference can find each plot later in often result in changes that we ask them for help. The Cooperative
in the way that each the season when the crop (and 10. Gather your data. If don’t anticipate. These changes Extension Service, the Natural
treatment is managed. weeds!) may have obliterated you are gathering harvest can be good, such as reduced pest Resources Conservation Service,
the original perimeter. You information, plan for enough pressure when soil organic matter the North Carolina Department of
4. Decide how many may want to use different time to keep harvested is increased, or bad, such as the Agriculture Agronomic Division,
replications you want. colored flags for different product from each treatment emergence of a non-target pest. crop consultants, non-government
Remember that the more treatments. in each block separate. Close and periodic observation organizations or university
replications that you have, can yield important information. researchers may all have some
the more confident you can 7. Outline your management 11. Evaluate the interest in assisting you in your
be in your results and conclu- practices (planting rates, experiment. You can get 4. Know how much confidence work. Just make sure that you ask
sions. After deciding how fertilizer, pesticide applica- help from Cooperative you can have in the results. Good as far in advance as possible to
large treatment plots will be, tions, tillage) for the crop Extension for a statistical design and numerous replications give them time to include it in
make a map showing how the and the dates you expect evaluation of the data can increase your confidence in their planning.
replications (each containing to complete each practice. you collect.
two treatment plots) will be This is the time, before you
laid out in the field. Within get into the field, to plan how
8 A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS 9STATISTICS 13. Take the square root of the Statistics allow us to calculate This LSD will be a smaller number
number from Step 12. the LSD, which tells us how large than the 95% LSD; thus it takes a
We can examine yield data to get answers Statistics take into account the chance that Record this value the difference between treatments smaller AVERAGE DIFFERENCE
to our research questions with some something may go unexpectedly 5% of the needs to be to: 1) account for possi- between treatments to pass the
mathematics called “statistics.” Impor- time. In other words, if we watch the two 14. Multiply the number in ble errors and random events, and test. However, in this case we
tantly, statistics confirm the answer to our teams play over and over and 95% of the time Step 13 by one of the numbers 2) to provide for a 95% certainty have only a 90% confidence level
research question, and tell us how reliable one team wins, we can be confident that they from Table 1 on page 12. The that the difference is real or (in that the difference is significant
that answer is. are the better team. The same thing is true with number you choose depends scientific terms) significant. The or real. Anything less than 90%
What purpose do statistics serve? Most of on-farm research: If we get the same result or on the number of blocks in the LSD is the smallest difference certainty is usually not considered
us have heard the expression “on any given answer to our research question 95% of the experiment and level of confidence between treatments that is mean- scientifically valid.
Sunday.” It has to do with professional football. time, statistics tell us that we have a valid, you want in the statistics (90% or ingful. If the AVERAGE DIFFER-
The “odds” may be that a superior team is scientifically “proven” answer. When we have 95% confident that the answer to ENCE between treatments (that 15. Compare the AVERAGE
going to beat an inferior team every time the six replications of the research experiment, your research question is correct). we calculated above) is smaller than DIFFERENCE (Column 5, step 6)
two teams play. But “on any given Sunday,” we can gather enough information to tell us This number (called the LSD the LSD, it’s too small a difference with the LSD calculated in Step 14.
or Least Significant to draw any conclusions from the In order for any differences
the opposite may happen. By luck, chance, or whether we can expect the answer to our
Difference) is: experiment. That is, it’s a small observed between treatments
circumstance, there is a certain probability question to be the same 95% of the time.
enough difference that it could to be significant, the AVERAGE
that a terrible team will beat a division leader. To calculate the experimental statistics,
The scientific method does have been caused by chance or DIFFERENCE must be greater
It might be a fluke play that brings defeat, or you will need a hand calculator that will give
not require 100% certainty for error. If the AVERAGE DIFFER- than the LSD. If the AVERAGE
an injury to the quarterback. you the square root of a number. Follow the
us to draw firm conclusions. ENCE is higher than the LSD, DIFFERENCE is greater than
Statistics allow us to account for flukes. steps outlined below. the LSD, then the
treatments are
significantly
Block Column 1 Column 2 Column 3 Block Column 4 Column 5 Column 6 Column 7
different. In the
Treatment 1 data Treatment 2 data Column 1 - Column 2 Copy from Column 3 Average Column 4 - Column 5 Column 6 x Column 6
case of our bean A
Begin the statistics by filling vs. bean B experi-
1 1
in the data for each Block. ment, whichever
2 2 variety has the higher
1. Put the data (for example, yield) average yield is the
for Treatment 1 in Column 1. more productive
3 3
2. Put the data for Treatment 2 variety. If it’s less
in Column 2. 4 4 than the LSD, we
say that there is no
3. In Column 3, subtract the 5 5 significant difference
number in Column 2 from the between treatments.
number in Column 1. (Some num- 6 6 In the latter case, we
bers in Column 3 may be negative, can’t say with scientif-
which is completely normal.) Average ic certainty that one
4. Calculate the AVERAGE of each Normally, we allow for random it’s too large a difference to be treatment (bean A) performs any
column. You do this by adding up 7. In Column 6, subtract the 9. Add all the numbers or chance errors and events to caused by chance or error alone. better yield-wise than the other
all the numbers in each column number in Column 5 from the in Column 7. create false results in an experi- We can be more than 95% certain treatment (bean B), even though
(positive and negative) and dividing number in Column 4. (Minus Record this value ment 5% of the time (1 out of that the difference between treat- data from all of the blocks of one
by the total number of blocks (in numbers can be tricky. Remember 20 times). For example, in a field, ments is real. treatment may be higher than in
10. Subtract 1 from the test bean A and bean B can be LSDs can be thought of as the other treatment.
this case we have six blocks). that (-5) - 5 = - 10. But also remem-
number of Blocks. expected to differ by an amount “passing grades.” To pass our Software programs are
ber that subtracting a minus number
5. Copy the numbers from Record this value as large as the LSD 5% of the time experimental test, the AVERAGE available that will do statistical
is the same as adding the number.
Column 3 into Column 4 just by chance. If we can predict DIFFERENCE must be at least calculations for you. Most spread-
For example, (-5) - (-5) is the same 11. Divide the number
(except the Column 3 AVERAGE from statistics that our data support as large as the LSD. If we can sheet programs will perform a
as (-5) + 5 and that this sum = 0.) from Step 9 by the number
DIFFERENCE). a conclusion that is 95% certain live with greater uncertainty, simple statistical analysis after
in Step 10.
6. Copy the Column 3 AVERAGE 8. In Column 7, square the num- Record this value (allowing for a 5% error), then we can calculate the LSD at the you type in your data. For more
into each and every row in ber in Column 6 (multiply it by we can be confident that we are 90% confidence level. It expresses information, contact your county
12. Divide the number giving people (ourselves included) the difference between Treatments Extension agent.
Column 5 (this will be the same itself). Note that a negative number
from Step 11 by the number good information when we 1 and 2 that we can expect to occur
number every row). “squared” is a positive number; for
of Blocks. report our results. by random chance 10% of the time.
example (-10) x (-10) = 100.
Record this value
10 A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS 11EXAMPLE
Block Column 1 Column 2 Column 3 Block Column 4 Column 5 Column 6 Column 7
STATISTICAL Treatment 1 data Treatment 2 data Column 1 - Column 2 Copy from Column 3 Average Column 4 - Column 5 Column 6 x Column 6
CALCULATION
1 190 165 25 1 25 15 10 100
2 210 180 30 2 30 15 15 225
3 180 185 -5 3 -5 15 -20 400
4 190 170 20 4 20 15 5 25
5 185 180 5 5 5 15 -10 100
6 195 180 15 6 15 15 0 0
Average 191.7 176.7 15
A grower wants to determine the 4. Next, she calculates the 8. In Column 7, she squares 11. She divides the number 13. Using her pocket calculator, 15. If the AVERAGE DIFFER-
effect of compost tea spray on AVERAGE of each Column. the number in Column 6 (mul- from Step 9 by the number she calculates the square root ENCE between the two treat-
plant pathogens of strawberry She does this by adding up all tiplies it by itself). Note that a from Step 10. This value of the number from Step 12. ments that she calculated
fruit. In a randomized and replicat- the numbers in each column negative number squared is a is 170 (850/5). The square root is 5.32. above is less than this LSD,
ed experimental field (with buffer (positive and negative) and positive number; for example there is no statistical signifi-
beds between treatment beds), dividing by the total number (-10) x (-10) = 100. 12. She divides the number 14. From Table 1 (below), she cant difference between the
she collects yield data from six of blocks (in this case we have from Step 11 by the number chooses a multiplier. Since she mean yields of each treatment.
beds that were sprayed with six blocks). 9. She adds the numbers in of Blocks. This value is 28.3 had six blocks and wants to be If the AVERAGE DIFFERENCE
compost tea (Treatment 1) and Column 7. This value is: 850. (170/6). 95% confident that her results is greater than the LSD, then
six beds that were not sprayed 5. Then she copies the are significant, she chooses 2.45. there is a statistical difference
with anything (Treatment 2). She numbers from Column 3 into 10. She subtracts 1 from She multiplies the number from between the two yields.
calculates her statistics as follows: Column 4 (except the Column 3 the number of Blocks. Step 13 by 2.45 to get
AVERAGE DIFFERENCE). This number is 6 - 1: 5. the Least Significant 16. When the grower
1. She puts the yield data for Difference (LSD). compares the AVERAGE
Treatment 1 in Column 1. 6. She copies the Column 3 This value is 13.04 DIFFERENCE from Column 5,
AVERAGE (15) into each and (5.32 x 2.45). Step 6 (15) with the LSD
2. She puts the yield data for every row in Column 5 (note calculated in Step 14 (13),
Treatment 2 in Column 2. that the number 15 is placed in she finds that the AVERAGE
every row in Column 5). DIFFERENCE is greater than
3. In Column 3, she subtracts Table 1. the LSD. She concludes with
the number in Column 2 from 7. In Column 6, the grower confidence that the compost
the number in Column 1. subtracts the number in # of 90% 95% tea treatment increased yield.
(Some numbers in Column 3 are Column 5 from the number Blocks Confidence Confidence
negative, which is completely in Column 4. (Minus numbers
normal.) can be tricky. Remember that 4 2.35 2.78
(-5) - 15 = -20.)
5 2.13 2.57
6 2.02 2.45
12 A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS A FIELD GUIDE for ON-FARM RESEARCH EXPERIMENTS 13Many thanks to Scott Marlow of the Rural
Advancement Foundation International-
USA (RAFI-USA) for his contributions to
“Helpful Hints” and his valuable com-
ments, suggestions, and review. Produced through the Cooperative Extension
Program at N.C. A&T State University
RAFI-USA provided funding for this publi-
cation through a grant from the North North Carolina A&T State University is committed to equality of
Carolina Rural Economic Development educational opportunity and does not discriminate against applicants,
Center, Inc., Agricultural Advancement students, or employees based on race, color, national origin, religion,
Consortium. Other funding for this work gender, age, or disability. Moreover, North Carolina A&T State University
was provided by the Southern Region is open to all people of all races and actively seeks to promote racial
Sustainable Agriculture Research and integration by recruiting and enrolling a large number of white students.
Education Program (SARE) and the North
Carolina Agricultural and Technical State
University Cooperative Extension Service.You can also read
