Tidy and plot meta-analyses from popular meta-analytic tools in R. Currently in early development.

Is this really all the entries for this manuscript? It seems rather large and unwieldy.

Could we have a description of the columns in this table added to the text?

I'm unclear on what "mean_treat_replicates" is. At first I though it was the number of replicates in the study but the fractional number doesn't support that and the col heading isn't clear.

Are we really excluding this many studies by only using those with 0.75m row spacing?

This code chunk suggests that, but the text prior does not mention this restriction:

@PaulMelloy, could you please create a simplified data file in .csv format for us to use that has the following columns and put it in the /data folder of the repository?

year (int)
location (char)
host_genotype (char)
row_spacing.m (double)
first_sign_disease (date)
fungicide_ai (char)
fungicide_application_1 (date)
fungicide_application_2 (date)
fungicide_application_3 (date)
total_fungicide (int)
grain_yield.t.ha (double)
grain_yield_error (double)
PM_final_severity (double)
PM_final_severity_error (double)

In cases where there are more than one control due to a fungicide application not being made, please average those rows into one value.

@emdelponte, are there other columns you would like to have or other suggestions?

I've updated my Rmd to cluster application times using Paul's new slim data.

Example in my branch for discussion. I've not yet committed any of this to Master aside from correcting the dates of application in 2018-19 season.

https://github.com/openplantpathology/Mungbean_PM/blob/Adam/Cluster%20Timing/MB_PM_data.md

Hello,
I have noticed that each time we impute the mean squares and run the metafor model I get different results.

I have added a code chunk following the analysis to show this at the following line.

I would really like to discuss how to deal with this so I can move forward when interpreting the results. What are your thought @adamhsparks and @emdelponte ?

Throwing a question out there,
I am probably missing a rookie error here
is there a way in R to specify how NAs are tested when asking if x == y?
ie if I want NA == NA to return TRUE is there a base function for that?

I wrote a function same() to handle them how I want the to be handled, but surely this is a problem others have had to?
Cheers

@adamhsparks can you check what I have done here.
Is this how you would define the dose categories?

Cheers

The Data wrangling vignette does not include all trials.

It is missing:

• Kingaroy 2012
• Emerald 2016
• Gatton 2012

Needs to be added

HigginsGreen is not in the bib file.

There is only a preprint link

There are quite a number of files in /R, many are related with weather data, which doesn't enter into this analysis. Any extraneous files should be removed before this repository is made public.

Also, the files should be checked for consistency and for any non-standard code, e.g. hard-coded paths as found in /R/Cleaning_historical_data.R, etc. to ensure that the workflow is reproducible

Duplicate chunk names, osset_Metafor-analysis

Hi @emdelponte

Do you know why there is an error in this code? I think it is with how the parameters are specified.

Error in forest.default(net_con) : Must specify either 'vi', 'sei', or ('ci.lb', 'ci.ub') pairs.

Cheers

I'm unable to reproduce this.

I keep getting this error.

r Error in rma.mv(yi = grain_yield.t.ha., V = vi, mods = ~spray_management * : Optimizer (optim) did not achieve convergence (convergence = 1).
from

See: https://github.com/openplantpathology/Mungbean_PM/blob/Adam/data/1902%20powdery%20mildew-Mungbean%20-%20Collated%20means.csv

I've split the Fungicide column into two cols, dose_ai_f1_g.ha and dose_ai_f2_g.ha so that we can have them properly formatted as double in R.

However, several of the rows, e.g., 87, 88, 95, 96, for dual mode-of-action fungicides only list one rate. This is inconsistent with other rows that correctly state the rate for both AIs.

The rate for fungicide AIs needs to be added for all fungicides. This is assuming we have the rate in the records.

From the text of the file itself.

This file is not complete and there are a number of things that need checking - Mean squares are only calculated for the triazoles and controls - If the years for each trial are input correctly - If the disease and yield errors are calculated and that they are shuffled correctly into the data frame - AUDPC needs to be checked for starting dates and if trials have been omitted - Check that what is generated from this file matches the main file and update - write.csv should have row.names = FALSE

If this isn't complete, then it needs to be checked and completed so that we can finish the analysis and then the paper that we've been working on.

It doesn't appear to be a documented vignette, rather it's several chunks of code just pasted in a Markdown file and run, to what purpose? How does this fit in the analysis workflow?

Also, I don't understand why it's not a part of the workflow itself if it is actually a part of the analysis that was performed.

In the commit log I found a message that the data file was now sourced from the shared OneDrive link. However, I cannot find any link in any of the Rmd files to the shared OneDrive link. Any data import lines point to the data/...xlsx file that no longer exists.

@paul Melloy, there's an issue in your function, same(), or at least how you've used it here:

It results in an error when I try to use it. It's because you're comparing something of length 1 with a vector of length 173 and so it takes only the first value of that vector, which may be what you want but it's not programmed correctly that I can tell.

master* > str(clusters[i, "cluster_1"])
 chr "Late"
master* > str(slim_PM_dat$cluster_1)
 chr [1:173] "Late" "Recommended" "Late" "Recommended" "Late" ...
master* 

This code chunk doesn't appear to do anything.

Is it broken or unnecessary?

Much of the code in this Rmd file is broken and needs to be fixed.

This should be looked into at a later date if there is a need for the information in this chapter

prop_YG is not found on line 506.

Additional meta-analysis models tested variables ...

Not sure what's supposed to be here?

See: https://github.com/openplantpathology/Mungbean_PM/blob/Adam/data/1902%20powdery%20mildew-Mungbean%20-%20Collated%20means.csv

The trial_design column says 2019 Hermitage were "RSP", I thought they were "RCB" as in other years?

See: https://github.com/openplantpathology/Mungbean_PM/blob/Adam/data/1902%20powdery%20mildew-Mungbean%20-%20Collated%20means.csv

This can wait, but the following chunk really needs more explanation as to what's being done.

@PaulMelloy, I'm going through the code in the book and am unable to determine the actual provenance of this file. This should be a read only file, but it looks like at some point it's overwritten during the book knitting?

How are there AUDPC and AUDPS values in this file? Those calculations should be done on-the-fly and if necessary results should be stored in the "cache" directory, which I created this morning in my branch I'm working on.

The data folder should be read only, we should not be writing back into it or any files contained inside of it.

Column headers for PM_MB_means were changed and the column names need to be corrected below accordingly

See

I don't find Powdery mildew paper#1.doc in the repository?

2012 and 2016 were both leap years, should we not account for this?

Adam to provide code review of data wrangling function to help ensure data integrity.

Line 14, PM_Mdat <- read.csv(here("cache/GYmeta_data.csv").

This file does not exist in the cache folder?

See: https://github.com/openplantpathology/Mungbean_PM/blob/Adam/data/1902%20powdery%20mildew-Mungbean%20-%20Collated%20means.csv

There are three Tilt treatments at this study site/year.

One, E_treament 3, has a single application, the other two, E_treatments 8 and 11 appear identical except for yield numbers?

The disease error is the same for both, both have only two applications and both were applied on the same dates.

What's the difference?

Line 224 fails.

Error in `[.data.frame`(slim_PM_dat, -rows_to_delete, ) : object 'rows_to_delete' not found

I have uploaded an abstract draft for submission to the Australasian Plant Pathology Society (APPS) conference. The deadline is tomorrow (25/7/19) at midnight.
Does anyone have any comments or edits?
The file can be found at ./Presentations/2019_APPS_Meta-PowderyMildewMungbean_Abstract.Rmd
Cheers

I've been looking at/working with the 02_weather.Rmd file and have some questions about it.

It has duplicate code in some places, and/or just doesn't run and it only has weather for a few locations, not all.

So, is this really necessary for our meta-analysis? I can't see what it really adds at this point and appears to have several issues in it's current form. If there's a clear reason for us to use it, I have some existing code for climograms that can clean it up and improve the figures, but I'm not convinced it's necessary to have for this work.

The text states that there appears to be no relationship between yield and AUDPC in the graph or linear model. But when I knit/run the linear model this is the result*:

Call:
lm(formula = grain_yield.t.ha. ~ AUDPC_m, data = PM_MB_means)

Residuals:
    Min      1Q  Median      3Q     Max 
-0.9620 -0.4077 -0.0783  0.3836  1.3954 

Coefficients:
            Estimate Std. Error t value            Pr(>|t|)    
(Intercept) 1.038082   0.051249   20.26 <0.0000000000000002 ***
AUDPC_m     0.001032   0.000395    2.61              0.0095 ** 
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Residual standard error: 0.537 on 314 degrees of freedom
  (39 observations deleted due to missingness)
Multiple R-squared:  0.0213,	Adjusted R-squared:  0.0181 
F-statistic: 6.82 on 1 and 314 DF,  p-value: 0.00945

Am I just confused about what's being referred to here? This is found here:

*In my branch I've removed the restriction to only look at row spacing of 0.75m. However, this model still indicates significance with or without this restriction.

In all but a few cases in this document, the references, e.g. Sparks2017 and Kelly2017a and AMAplanting, should be peer-reviewed.

Currently in the .github/workflows folder the file deploy_bookdown.yml specifies using Cecilapp/GitHub-Pages-deploy@v2

This is the old version and should be updated to version 3 as per the readme on
https://github.com/Cecilapp/GitHub-Pages-deploy

Line

Several places seem to reference the use of netMeta but I can't find any analysis actually carried out with it.