[R] Odd Results when generating predictions with nnet function

Jeff Newmiller jdnewm|| @end|ng |rom dcn@d@v|@@c@@u@
Wed Sep 2 17:01:22 CEST 2020 

Why would you expect raising y_pred to the 0.5 to "backtransform" a model sqrt(y)~x? Wouldn't you raise to the 2?

Why would you "backtransform" x in such a model if it were never transformed in the first place? Dr Maechler did not suggest that.

And why are you mentioning some random unspecified video on Youtube? That does not enlighten anyone here, apparently including you. Please reference package documentation, and/or reproduce the analysis discussed in that video to provide a contrasting (or supporting) point with the example you gave.


On September 2, 2020 7:21:58 AM PDT, Paul Bernal <paulbernal07 using gmail.com> wrote:
>Dear Dr. Martin and Dr. Peter,
>
>Hope you are doing well. Thank you for your kind feedback. I also tried
>fitting the nnet using y ~ x, but the model kept on generating odd
>predictions. If I understand correctly, from what Dr. Martin said, it
>would
>be a good idea to try modeling sqrt(y) ~ x and then backtransform
>raising
>both y and x to 0.5?
>
>I was looking at a video where the guy modeled count data without doing
>any
>kind of transformation and didn't get odd results, which is rather
>extrange.
>
>Cheers,
>
>Paul
>
>
>
>El mié., 2 sept. 2020 a las 2:37, Martin Maechler (<
>maechler using stat.math.ethz.ch>) escribió:
>
>> >>>>> peter dalgaard
>> >>>>>     on Wed, 2 Sep 2020 08:41:09 +0200 writes:
>>
>>     > Generically, nnet(a$y ~ a$x, a ...) should be nnet(y ~ x,
>>     > data=a, ...) otherwise predict will go looking for a$x, no
>>     > matter what is in xnew.
>>
>>     > But more importantly, nnet() is a _classifier_,
>>     > so the LHS should be a class, not a numeric variable.
>>
>>     > -pd
>>
>> Well, nnet() can be used for both classification *and* regression,
>> which is quite clear from the MASS book, but indeed, not from
>> its help page, which indeed mentions one formula  'class ~ ...'
>> and then only has classification examples.
>>
>> So, indeed, the  ?nnet  help page could improved.
>>
>> In his case, y are counts,  so  John Tukey's good old
>> "first aid transformation" principle would suggest to model
>>
>> sqrt(y) ~ ..   in a *regression* model which nnet() can do.
>>
>> Martin Maechler
>> ETH Zurich  and  R Core team
>>
>>
>>
>>     >> On 1 Sep 2020, at 22:19 , Paul Bernal
>>     >> <paulbernal07 using gmail.com> wrote:
>>     >>
>>     >> Dear friends,
>>     >>
>>     >> Hope you are all doing well. I am currently using R
>>     >> version 4.0.2 and working with the nnet package.
>>     >>
>>     >> My dataframe consists of three columns, FECHA which is
>>     >> the date, x, which is a sequence from 1 to 159, and y,
>>     >> which is the number of covid cases (I am also providing
>>     >> the dput for this data frame below).
>>     >>
>>     >> I tried fitting a neural net model using the following
>>     >> code:
>>     >>
>>     >> xnew = 1:159 Fit <- nnet(a$y ~ a$x, a, size = 5, maxit =
>>     >> 1000, lineout = T, decay = 0.001)
>>     >>
>>     >> Finally, I attempted to generate predictions with the
>>     >> following code:
>>     >>
>>     >> predictions <- predict(Fit, newdata = list(x = xnew),
>>     >> type = "raw")
>>     >>
>>     >> But obtained extremely odd results: As you can see,
>>     >> instead of obtaining numbers, more or less in the range
>>     >> of the last observations of a$y, I end up getting a bunch
>>     >> of 1s, which doesn´t make any sense (if anyone could help
>>     >> me understand what could be causing this):
>>     >> dput(predictions) structure(c(1, 1, 1, 1, 1, 1, 1, 1, 1,
>>     >> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
>>     >> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
>>     >> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
>>     >> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
>>     >> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
>>     >> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
>>     >> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
>>     >> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1), .Dim
>>     >> = c(159L, 1L), .Dimnames = list(c("1", "2", "3", "4",
>>     >> "5", "6", "7", "8", "9", "10", "11", "12", "13", "14",
>>     >> "15", "16", "17", "18", "19", "20", "21", "22", "23",
>>     >> "24", "25", "26", "27", "28", "29", "30", "31", "32",
>>     >> "33", "34", "35", "36", "37", "38", "39", "40", "41",
>>     >> "42", "43", "44", "45", "46", "47", "48", "49", "50",
>>     >> "51", "52", "53", "54", "55", "56", "57", "58", "59",
>>     >> "60", "61", "62", "63", "64", "65", "66", "67", "68",
>>     >> "69", "70", "71", "72", "73", "74", "75", "76", "77",
>>     >> "78", "79", "80", "81", "82", "83", "84", "85", "86",
>>     >> "87", "88", "89", "90", "91", "92", "93", "94", "95",
>>     >> "96", "97", "98", "99", "100", "101", "102", "103",
>>     >> "104", "105", "106", "107", "108", "109", "110", "111",
>>     >> "112", "113", "114", "115", "116", "117", "118", "119",
>>     >> "120", "121", "122", "123", "124", "125", "126", "127",
>>     >> "128", "129", "130", "131", "132", "133", "134", "135",
>>     >> "136", "137", "138", "139", "140", "141", "142", "143",
>>     >> "144", "145", "146", "147", "148", "149", "150", "151",
>>     >> "152", "153", "154", "155", "156", "157", "158", "159"),
>>     >> NULL))
>>     >>
>>     >> head(a) FECHA x y 1 2020-03-09 1 1 2 2020-03-10 2 8 3
>>     >> 2020-03-11 3 14 4 2020-03-12 4 27 5 2020-03-13 5 36 6
>>     >> 2020-03-14 6 43
>>     >>
>>     >> dput(a) structure(list(FECHA = structure(c(18330, 18331,
>>     >> 18332, 18333, 18334, 18335, 18336, 18337, 18338, 18339,
>>     >> 18340, 18341, 18342, 18343, 18344, 18345, 18346, 18347,
>>     >> 18348, 18349, 18350, 18351, 18352, 18353, 18354, 18355,
>>     >> 18356, 18357, 18358, 18359, 18360, 18361, 18362, 18363,
>>     >> 18364, 18365, 18366, 18367, 18368, 18369, 18370, 18371,
>>     >> 18372, 18373, 18374, 18375, 18376, 18377, 18378, 18379,
>>     >> 18380, 18381, 18382, 18383, 18384, 18385, 18386, 18387,
>>     >> 18388, 18389, 18390, 18391, 18392, 18393, 18394, 18395,
>>     >> 18396, 18397, 18398, 18399, 18400, 18401, 18402, 18403,
>>     >> 18404, 18405, 18406, 18407, 18408, 18409, 18410, 18411,
>>     >> 18412, 18413, 18414, 18415, 18416, 18417, 18418, 18419,
>>     >> 18420, 18421, 18422, 18423, 18424, 18425, 18426, 18427,
>>     >> 18428, 18429, 18430, 18431, 18432, 18433, 18434, 18435,
>>     >> 18436, 18437, 18438, 18439, 18440, 18441, 18442, 18443,
>>     >> 18444, 18445, 18446, 18447, 18448, 18449, 18450, 18451,
>>     >> 18452, 18453, 18454, 18455, 18456, 18457, 18458, 18459,
>>     >> 18460, 18461, 18462, 18463, 18464, 18465, 18466, 18467,
>>     >> 18468, 18469, 18470, 18471, 18472, 18473, 18474, 18475,
>>     >> 18476, 18477, 18478, 18479, 18480, 18481, 18482, 18483,
>>     >> 18484, 18485, 18486, 18487, 18488), class = "Date"), x =
>>     >> 1:159, y = c(1, 8, 14, 27, 36, 43, 55, 69, 86, 109, 137,
>>     >> 200, 245, 313, 345, 443, 558, 674, 786, 901, 989, 1075,
>>     >> 1181, 1317, 1475, 1673, 1801, 1988, 2100, 2249, 2528,
>>     >> 2752, 2974, 3234, 3400, 3472, 3574, 3751, 4016, 4210,
>>     >> 4273, 4467, 4658, 4821, 4992, 5166, 5338, 5538, 5779,
>>     >> 6021, 6200, 6378, 6532, 6720, 7090, 7197, 7387, 7523,
>>     >> 7731, 7868, 8070, 8282, 8448, 8616, 8783, 8944, 9118,
>>     >> 9268, 9449, 9606, 9726, 9867, 9977, 10116, 10267, 10577,
>>     >> 10926, 11183, 11447, 11728, 12131, 12531, 13015, 13463,
>>     >> 13837, 14095, 14609, 15044, 15463, 16004, 16425, 16854,
>>     >> 17233, 17889, 18586, 19211, 20059, 20686, 21422, 21962,
>>     >> 22597, 23351, 24274, 25222, 26030, 26752, 27314, 28030,
>>     >> 29037, 29905, 30658, 31686, 32785, 33550, 34463, 35237,
>>     >> 35995, 36983, 38149, 39334, 40291, 41251, 42216, 43257,
>>     >> 44352, 45633, 47177, 48096, 49243, 50373, 51408, 52261,
>>     >> 53468, 54426, 55153, 55906, 56817, 57993, 58864, 60296,
>>     >> 61442, 62223, 63269, 64191, 65256, 66383, 67453, 68456,
>>     >> 69424, 70231, 71418, 72560, 73651, 74492, 75394, 76464,
>>     >> 77377, 78446, 79402)), row.names = c(NA, 159L), class =
>>     >> "data.frame") Any help and/or guidance will be greatly
>>     >> appreciated,
>>     >>
>>     >> Cheers,
>>     >>
>>     >> Paul
>>     >>
>>     >> [[alternative HTML version deleted]]
>>     >>
>>     >> ______________________________________________
>>     >> R-help using r-project.org mailing list -- To UNSUBSCRIBE and
>>     >> more, see https://stat.ethz.ch/mailman/listinfo/r-help
>>     >> PLEASE do read the posting guide
>>     >> http://www.R-project.org/posting-guide.html and provide
>>     >> commented, minimal, self-contained, reproducible code.
>>
>>     > --
>>     > Peter Dalgaard, Professor, Center for Statistics,
>>     > Copenhagen Business School Solbjerg Plads 3, 2000
>>     > Frederiksberg, Denmark Phone: (+45)38153501 Office: A 4.23
>>     > Email: pd.mes using cbs.dk Priv: PDalgd using gmail.com
>>
>>     > ______________________________________________
>>     > R-help using r-project.org mailing list -- To UNSUBSCRIBE and
>>     > more, see https://stat.ethz.ch/mailman/listinfo/r-help
>>     > PLEASE do read the posting guide
>>     > http://www.R-project.org/posting-guide.html and provide
>>     > commented, minimal, self-contained, reproducible code.
>>
>
>	[[alternative HTML version deleted]]
>
>______________________________________________
>R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
>https://stat.ethz.ch/mailman/listinfo/r-help
>PLEASE do read the posting guide
>http://www.R-project.org/posting-guide.html
>and provide commented, minimal, self-contained, reproducible code.

-- 
Sent from my phone. Please excuse my brevity.

More information about the R-help mailing list