0

I'd like to run a prediction across different alphas using the glmnet package.

y is a (14x1) dependent variable, x is a series (14x19) of independent variables.

Dput data below.

x <- Macro[1:13,3:21]
x <- as.matrix(x)   
y <- Macro[1:13,2:2]
y <- as.matrix(y)
t <- Macro[14:14,3:21]
t <- as.matrix(t)

Ideally I'm looking to return a different prediction for each alpha:

for (alpha in c(0,.1,.3,.5,.7,.9,1)) {
cv.fit=cv.glmnet(x,y, alpha = alpha, lambda = NULL, nfolds = 5)
min <- cv.fit$lambda.min
fit <- cv.glmnet(x, y, alpha=alpha, lambda = cv.fit$lambda.min, nfolds=5)
predict(fit ,t, s = min)
}

But this returns an error:

Error in cv.glmnet(x, y, alpha = alpha, lambda = cv.fit$lambda.min, nfolds = 5) : 
Need more than one value of lambda for cv.glmnet

Note the code:

 for (alpha in c(0,.1,.3,.5,.7,.9,1)) {
  fit <- cv.glmnet(x,y, alpha=alpha, nfolds = 5)
}

works fine.

How can I fix my code? I really need to execute the predict function in the loop for the alphas, not just the fit.

dput(x)
structure(c(4.652228652, 4.452166281, 5.553326349, 4.188964308, 
5.012729352, 4.538928371, 5.638139338, 3.925012902, 4.682906379, 
4.660168251, 5.952803094, 4.721206697, 5.758055685, 1.903978779, 
1.837148025, 2.027326755, 1.686266994, 2.017370721, 1.773818155, 
1.937293325, 1.647630534, 1.749950342, 1.730985306, 2.263447785, 
1.98126732, 2.434642854, 2.481570287, 2.33258972, 2.912199168, 
2.110299378, 2.595402032, 2.322950029, 2.842142129, 1.979896372, 
2.319502793, 2.359240313, 2.937856446, 2.399225856, 2.827140745, 
5.082707128, 4.712472353, 5.516298089, 4.168246822, 5.267110329, 
4.833249636, 5.828918236, 4.140403816, 5.28575776, 4.955121909, 
6.404571778, 4.880203713, 6.640952257, 2.320933796, 2.120317491, 
2.593030466, 1.969162713, 2.195998477, 2.198837636, 2.6051228, 
1.87366517, 2.1863434, 2.112724392, 2.625023126, 2.069334825, 
2.328814677, 2.468942531, 2.46345831, 2.99338684, 2.349608577, 
2.568479669, 2.600346713, 3.056925547, 2.129869136, 2.449644735, 
2.352858179, 2.924043472, 2.26104673, 2.411660085, 1.692902733, 
1.541408483, 1.947110878, 1.520008357, 1.671217322, 1.660831673, 
1.992684923, 1.40046815, 1.644093122, 1.581270255, 1.937395811, 
1.519664787, 1.77809053, 2.130588433, 1.928360314, 2.501628649, 
1.930166962, 2.19632807, 2.024905776, 2.667980643, 1.837998822, 
2.346063091, 2.137170693, 2.745812681, 2.111553082, 2.352218846, 
0.040721902, 0.039608433, 0.055438722, 0.037340065, 0.044442281, 
0.041467602, 0.056821845, 0.036648017, 0.044310579, 0.042240444, 
0.060200942, 0.03989574, 0.05027253, 0.019096759, 0.01877026, 
0.024778923, 0.016531124, 0.019846859, 0.017973319, 0.024291258, 
0.017512498, 0.019927366, 0.018047551, 0.027240891, 0.019572431, 
0.024967503, 0.021136035, 0.019737199, 0.029282295, 0.018555507, 
0.022862438, 0.020947225, 0.029544409, 0.018154018, 0.021944567, 
0.021016969, 0.030287821, 0.020090802, 0.024800349, 0.044050512, 
0.041113277, 0.054492472, 0.038055331, 0.047872178, 0.044639163, 
0.059491776, 0.039308651, 0.04955884, 0.047517015, 0.064765233, 
0.043519587, 0.060235515, 0.029467808, 0.027428719, 0.036827997, 
0.02638519, 0.029086009, 0.0290847, 0.037315356, 0.025529541, 
0.029876725, 0.028196018, 0.037560361, 0.026557948, 0.030220349, 
0.03750512, 0.03653175, 0.048462325, 0.036646121, 0.038238758, 
0.038760921, 0.049783835, 0.033023068, 0.037417803, 0.035506876, 
0.046763041, 0.033261873, 0.034905163, 0.022926791, 0.021173154, 
0.029388903, 0.021020651, 0.023448143, 0.023490222, 0.030767307, 
0.020584604, 0.024119676, 0.022463317, 0.029776277, 0.020834843, 
0.024352043, 0.025042587, 0.023646145, 0.032299653, 0.024133764, 
0.026572026, 0.024691929, 0.034760273, 0.023298189, 0.02887728, 
0.026780364, 0.035738399, 0.024864652, 0.027768462, 1, 0, 0, 
0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 
0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0), .Dim = c(13L, 19L
), .Dimnames = list(c("1", "2", "3", "4", "5", "6", "7", "8", 
"9", "10", "11", "12", "13"), c("c1372", "c5244", "c5640", "c6164", 
"b1372", "b5244", "b5640", "b6164", "v1372", "v5244", "v5640", 
"v6164", "bv1372", "bv5244", "bv5640", "bv6164", "s1", "s2", 
"s3")))
> dput(y)
structure(c(867.46, 825.258, 1013.406, 762.577, 890.568, 862.491, 
1030.2, 761.2, 872.93, 892.77, 1089.12, 855.69, 992.454), .Dim = c(13L, 
1L))
> dput(t)
structure(c(5.154502875, 2.04158867, 2.717400485, 6.035902763, 
2.22687594, 2.513986337, 1.77744695, 2.146230648, 0.046828898, 
0.021535376, 0.024242411, 0.056390357, 0.029433671, 0.03603116, 
0.024883907, 0.026647629, 0, 1, 0), .Dim = c(1L, 19L), .Dimnames = list(
    "14", c("c1372", "c5244", "c5640", "c6164", "b1372", "b5244", 
    "b5640", "b6164", "v1372", "v5244", "v5640", "v6164", "bv1372", 
    "bv5244", "bv5640", "bv6164", "s1", "s2", "s3")))
ZJAY
  • 2,517
  • 9
  • 32
  • 51
  • Can you make your problem reproducible by including sample data? However, error message suggests that you need "more than one value of lambda for cv.glmnet". Try changing this part of your code: `lambda = cv.fit$lambda.min`. – markus Aug 14 '18 at 19:18
  • Thanks. For each loop, there should only be one value of cv.fit$lambda.min - any idea of a fix? – ZJAY Aug 14 '18 at 20:47
  • After reading a bit in `?cv.glmnet` I would recommend changing the code to be `predict(cv.fit, newx = t, s = "lambda.min")` without defining `min` and `fit`. – markus Aug 14 '18 at 20:54
  • 1
    Use `dput(your_data)` to share your data. No pictures please. – markus Aug 14 '18 at 20:55
  • @markus - unfortunately I got the same error with your edits: Error in glmnet::cv.glmnet(x, y, ...) : Need more than one value of lambda for cv.glmnet – ZJAY Aug 14 '18 at 21:10

1 Answers1

0

Just put

predict(cv.fit ,t, s = 'lambda.min')

instead of

fit <- cv.glmnet(x, y, alpha=alpha, lambda = cv.fit$lambda.min, nfolds=5)
predict(fit ,t, s = min)

There is no need to fit the model again

Evgeny Ivanov
  • 1
  • 1
  • This produced the error Option grouped=FALSE enforced in cv.glmnet, since < 3 observations per fold - how do I access the predictions? – ZJAY Sep 09 '18 at 00:21