Clarification on "Jtrans" package

[SyamkumarR]

Recently, I came across an R-package "jtrans" and found to be very useful. I would like to clarify the following doubts regarding the package.

1. How do I back transform the data?
2. I have a non-normal dataset which I tried to transform using the package, but didnt work. Can you suggest me a solution?

The codes are given below:

read.table(textConnection("treat repl effect1 transformed
ctrl 1 4.5568 -0.94855895
ctrl 2 7.2874 -0.61241869
ctrl 3 2.0557 -1.5307805
ctrl 4 0.8671 -2.24734323
ctrl 5 2.0416 -1.53599825
treat1 1 13.9265 -0.13786884
treat1 2 13.098667 -0.18394011
treat1 3 14.4454 -0.11019863
treat1 4 15.944 -0.03484184
treat1 5 15.307182 -0.06608971
treat2 1 37.434 0.69328341
treat2 2 37.175889 0.6864524
treat2 3 22.9898 0.25623886
treat2 4 14.8152 -0.0910007
treat2 5 14.589091 -0.10268909
treat3 1 10.0914 -0.37676977
treat3 2 14.581556 -0.1030813
treat3 3 2.1069 -1.51216859
treat3 4 1.7222 -1.6666435
treat3 5 2.516182 -1.37945895
treat4 1 28.5665 0.44154336
treat4 2 24.951333 0.32471788
treat4 3 31.5415 0.53054774
treat4 4 43.0154 0.83616312
treat4 5 30.381909 0.49651044
treat5 1 21.1001 0.18601383
treat5 2 14.885111 -0.08741789
treat5 3 17.2256 0.02497832
treat5 4 13.227 -0.17663297
treat5 5 12.561 -0.21526148
treat6 1 9.1758 -0.44606229
treat6 2 6.126222 -0.73683883
treat6 3 7.7394 -0.5691373
treat6 4 7.8217 -0.56151922
treat6 5 8.132455 -0.53343019
treat7 1 94.2584 2.3052373
treat7 2 98.629556 2.55639605
treat7 3 99.6842 2.63047148
treat7 4 90.0177 2.11485905
treat7 5 80.996182 1.79815736
treat8 1 43.508 0.84840624
treat8 2 46.841889 0.93013131
treat8 3 46.3398 0.91793466
treat8 4 42.6801 0.82780097
treat8 5 42.992273 0.8355871
treat9 1 56.5082 1.16099219
treat9 2 57.110111 1.17530181
treat9 3 52.3417 1.06202199
treat9 4 44.8252 0.88091602
treat9 5 61.326909 1.27612697
treat10 1 1.5229 -1.76342652
treat10 2 2.020444 -1.54390408
treat10 3 0.8935 -2.21915988
treat10 4 0.519 -2.81626084
treat10 5 0.907 -2.20520354
treat11 1 8.9524 -0.4639497
treat11 2 9.410111 -0.42773525
treat11 3 10.884 -0.3213404
treat11 4 7.9897 -0.54620392
treat11 5 8.716091 -0.4833343
"),header=T)->dat

jt<- jtrans(dat$effect1)
plot(density(dat$effect1))
plot(density(jt$transformed))
qqnorm(jt$transformed)
qqline(jt$transformed)
jtrans(md5$root, test = "ad.test")
predict(jt, dat$effect1)

dat$transformed<-jt$transformed
aov(transformedtreat,data=dat)->m1
shapiro.test(resid(m1))
ad.test(resid(m1))
gls(transformedtreat,data=dat)->m1
shapiro.test(resid(m1))
ad.test(resid(m1))

[wangyuchen]

Transform it back is a good idea. I'll update it with this feature. Now exporting a LaTeX equation using transeq(jt) is the best it can do, you can easily transform it back based on that.

As for your data, I did the transformation and both shapiro.test and nortest::ad.test give a non-significant p-value. Their density plots all look quite normal. Sometimes the transformation doesn't work when the original distribution is bimodal, but your data doesn't look like that.

[wangyuchen]

@SyamkumarR What exactly is the use case of transforming it back? If it was on the original data, you should have it already; if it was on another normal data, I don't see why you want to do that, and the transformation may not apply, i.e. the target distribution may not be your original distribution.

[SyamkumarR]

Hi,
It is customary to have the data back-transformed when reporting.
Otherwise, it becomes difficult to interpret. For example, it is difficult
to interpret the mean "leaf area" on a log scale (log transformed).

Syamkumar R
Research Scholar,
SES, CUSAT

On Fri, Nov 4, 2016 at 7:47 PM, Yuchen Wang [email protected]
wrote:

@SyamkumarR https://github.com/SyamkumarR What exactly is the use case
of transforming it back? If it was on the original data, you should have it
already; if it was on another normal data, I don't see why you want to do
that, and the transformation may not apply, i.e. the target distribution
may not be your original distribution.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#1 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AWB8KGH4jpeEso5nNaA5TAXV-ltih89Rks5q6z6NgaJpZM4KiWdB
.