bose<-coef(lasso.mod, x=lassotemp$x, y=lassotemp$y, weights=lassotemp$weight, s=cv.out$lambda.1se, exact=TRUE)[-1]

fixedLassoInf(lassotemp$x,lassotemp$y, bose,lambda=cv.out$lambda.1se*8)

fixedLassoInf(lassotemp$x,lassotemp$y, bose,lambda=cv.out$lambda.1se*8)

I get the following result:

Standard deviation of noise (specified or estimated) sigma = 0.480

Testing results at lambda = 0.002, with alpha = 0.100

Var Coef Z-score P-value LowConfPt UpConfPt LowTailArea UpTailArea

1 0.000 3.229 1 -Inf 0.000 0 0

3 0.091 3.579 1 -Inf -2.552 0 NaN

4 -0.029 -1.988 1 1.434 Inf NaN 0

5 -0.039 -3.127 1 1.232 Inf 0 0

6 -0.078 -2.841 NaN 2.735 Inf NaN 0

7 0.104 1.907 1 -Inf -5.448 0 NaN

8 0.101 3.575 1 -Inf -2.839 0 NaN

10 0.000 -1.438 1 0.000 Inf NaN 0

11 -0.077 -1.502 1 5.140 Inf NaN 0

12 0.126 2.713 1 -Inf -4.644 0 NaN

13 0.007 2.414 1 -Inf -0.306 0 NaN

14 -0.002 -1.270 1 0.145 Inf NaN 0

15 0.000 6.834 1 -Inf -0.003 0 NaN

16 0.116 3.741 1 -Inf -3.104 0 NaN

17 0.000 0.007 1 -Inf -1.487 0 0

19 0.005 0.056 1 -Inf -8.744 0 NaN

20 0.070 2.837 1 -Inf -2.458 0 NaN

21 0.050 1.213 1 -Inf -4.081 0 NaN

22 0.005 0.148 1 -Inf -3.287 0 NaN

23 0.021 0.656 1 -Inf -3.181 0 NaN

24 -0.359 -1.767 1 20.348 Inf 0 0

25 -0.074 -1.549 1 4.772 Inf NaN 0

26 -0.206 -8.226 1 2.508 Inf NaN 0

27 -0.063 -1.263 1 5.018 Inf NaN 0

28 0.151 4.101 1 -Inf -3.693 0 NaN

29 0.333 5.950 1 -Inf -5.596 0 NaN

31 -0.028 -0.776 1 3.572 Inf NaN 0

32 -0.116 -3.713 1 3.116 Inf 0 0

33 -0.112 -3.777 1 2.959 Inf NaN 0

34 0.183 6.341 1 -Inf -2.882 0 NaN

35 -0.119 -3.529 1 3.373 Inf NaN 0

36 0.053 1.127 1 -Inf -4.692 0 NaN

37 0.074 1.683 1 -Inf -4.375 0 NaN

38 0.043 0.839 1 -Inf -5.115 0 NaN

39 0.122 2.419 1 -Inf -5.033 0 NaN

40 0.202 2.678 1 -Inf -7.543 0 NaN

41 0.071 1.009 1 -Inf -7.064 0 NaN

42 -0.174 -3.563 1 4.871 Inf NaN 0

43 -0.057 -1.012 1 5.616 Inf NaN 0

44 0.120 2.237 1 -Inf -5.346 0 NaN

45 -0.115 -2.240 1 5.149 Inf NaN 0

46 -0.155 -2.056 1 7.523 Inf NaN 0

47 -0.048 -0.845 1 5.700 Inf NaN 0

49 0.039 0.725 1 -Inf -5.437 0 NaN

50 -0.403 -8.360 1 4.823 Inf 0 0

51 -0.150 -2.285 1 6.584 Inf NaN 0

52 -0.142 -1.978 1 7.205 Inf NaN 0

53 -0.071 -1.254 1 5.693 Inf NaN 0

54 -0.139 -2.336 1 5.963 Inf NaN 0

55 -0.260 -2.739 1 9.506 Inf NaN 0

56 -0.598 -7.131 1 8.392 Inf NaN 0

57 -0.253 -1.899 1 13.332 Inf 0 0

58 -0.544 -7.194 1 7.565 Inf NaN 0

59 -0.392 -4.033 1 9.715 Inf NaN 0

61 -0.140 -1.679 1 8.331 Inf NaN 0

62 -0.407 -3.422 1 11.906 Inf NaN 0

63 -0.259 -4.187 1 6.177 Inf NaN 0

64 -0.416 -5.548 1 7.491 Inf NaN 0

65 -0.595 -8.929 NaN 6.665 Inf NaN 0

66 -0.169 -3.375 1 5.003 Inf NaN 0

67 -0.359 -7.107 1 5.057 Inf NaN 0

68 -0.310 -6.923 1 4.485 Inf NaN 0

69 -0.456 -8.973 1 5.087 Inf 0 0

70 0.019 3.179 1 -Inf -0.583 0 NaN

71 -0.001 -4.969 1 0.011 Inf NaN 0

72 -0.012 -1.471 1 0.816 Inf NaN 0

73 0.000 -0.252 1 0.052 Inf NaN 0

74 0.040 4.468 1 -Inf -0.888 0 0

Note: coefficients shown are partial regression coefficients

Warning messages:

1: In fixedLassoInf(lassotemp$x, lassotemp$y, bose, lambda = cv.out$lambda.min, :

Solution beta does not satisfy the KKT conditions (to within specified tolerances)

2: In fixedLassoInf(lassotemp$x, lassotemp$y, bose, lambda = cv.out$lambda.min, :

Solution beta does not satisfy the KKT conditions (to within specified tolerances). You might try rerunning glmnet with a lower setting of the 'thresh' parameter, for a more accurate convergence.

----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

First question: is it correct, to multiply the lambda value by 8 when calling fixedLassoInf? The documentation for fixedlassoinf says to divide the lambda value when getting the coef by 8, but then I do not get the coefficients I am looking for. I found some slides from some university professor online, who did not divide or multiply anything, thus I am not sure whether my approach is correct?

Next: I found that for Lambda values of 0.05 fixedLassoInf provides reasonable results. But not for small Lambda values, as my chosen 1se lambda of 0.002. Can anybody explain why that is?

Finally: I also found that for a large tolerance value for beta (tol.beta = 1) fixedlassoinf gives reasonable results for the small lambda values, can I just increase the tolerance level, or is that simply wrong?

[PS. There are no missing values in the data, as has been the source of similar problems for others]