[R] Problem with random numbers/seed

[Uwe Ligges]

On 21.06.2019 19:09, Steven Yen wrote:
> Now I see that results were replicated but running RNGversion
> I get a warning message. Isn't there a way to do this clean?

Well, the old RNG is known to be not optimal, hence we give a warning if 
you choose the old one. This is the clean way if you really want to get 
the old (slightly buggy) behaviour.

Best,
Uwe Ligges

> 
>> RNGversion("3.5.3") Warning message: In RNGkind("Mersenne-Twister",
> "Inversion", "Rounding") : non-uniform 'Rounding' sampler used
> 
> 
> On 6/22/2019 1:03 AM, Uwe Ligges wrote:
>> What does not work?
>>
>>
>> For me it works under R-3.6.0:
>>
>>   x<-1:500
>>   set.seed(12345671)
>>   j<-sample(1:length(x),size=60); y<-x[j]
>>   summary(j)
>>
>>   RNGversion("3.5.3")
>>   set.seed(12345671)
>>   j<-sample(1:length(x),size=60); y<-x[j]
>>   summary(j)
>>
>>
>> Now I get the results you got udner the old R.
>>
>> Best,
>> Uwe Ligges
>>
>>
>>
>> On 21.06.2019 18:39, Steven Yen wrote:
>>> Thanks. Somewhat of a mystery. The older version I had was
>>> R-3.5.3patched.
>>> I cannot get the RNGversion command to run. Can you help? Thanks.
>>>
>>> On 6/22/2019 12:25 AM, Uwe Ligges wrote:
>>>> See the NEWS, the RNG has been changed, use RNGversion
>>>>
>>>> On 21.06.2019 18:10, Steven Yen wrote:
>>>>> Dear all,
>>>>> I did all this work with older R (R-3.5.3.patched and older)
>>>>> but now with R3.6 I cannot replicate the results.
>>>>> Below I sample 60 observations from 1:500 using the sample command
>>>>> with
>>>>> a random seed of 123. I get different results. Advice appreciated.
>>>>> Steven Yen
>>>>>
>>>>>    > # Run under R-3.6.0
>>>>>    > x<-1:500
>>>>>    > set.seed(12345671)
>>>>>    > j<-sample(1:length(x),size=60); y<-x[j]
>>>>>    > summary(y)
>>>>>       Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
>>>>>       26.0   134.2   240.0   249.8   368.0   500.0
>>>>>
>>>>>    > # Run under R-3.5.3.patched
>>>>>    > x<-1:500
>>>>>    > set.seed(12345671)
>>>>>    > j<-sample(1:length(x),size=60); y<-x[j]
>>>>>    > summary(y)
>>>>>       Min. 1st Qu.  Median    Mean 3rd Qu.    Max.
>>>>>        9.0   122.2   205.0   236.1   364.2   493.0
>>>>>
>>>>
>>>>
>>>> Under R-3.6.0  use, e.g.
>>>> RNGversion("3.5.2")
>>>> to get reproducible results from the older RNG.
>>>>
>>>> Best,
>>>> Uwe Ligges
>>>>
>>>>
>>>
>>> -- 
>>> styen using ntu.edu.tw  (S.T. Yen)
>>>
>>
>