1
$\begingroup$

I have 2 distributions, which are not normally distributed. I would like to add and subtract them. I know it will work if those 2 distributions are normally distributed.

I have tried to use fitdistr in R language to bring those distributions into normal distributions. My problem is the new normal distributions have negative values, which in my case is not permitted.

So, I have 2 questions: 1. Is there a way to combine (add / substract) distribution which are not normally distributed? 2. Can make a normal distribution based on non-normal one without having negative values?

These are the distribution I would like to add and subtract:

dist_a <- 304.123514   7.765153  31.564579  14.833717   8.465059   6.258208 110.321460 448.445840  92.832490  81.560523  10.732478   12.215746  52.976102  49.655802 177.337262  54.131492   4.244143   6.822287 245.103775  14.207249  26.560730  78.116001 136.886305  14.517104 139.871747  15.825593 580.965041  12.215746  11.450119  70.127765   5.865972  34.409630   1.714944 131.105225  92.832490 112.727532  54.131492   9.429312 125.568296   6.676671  46.543603  36.710475  38.329222 142.922301 5.740768  25.993814   1.474552  18.012749  22.837576  24.896024  74.816950  35.926921  33.675186   6.394697   6.971079  107.966743  33.675186  28.336747  30.890860  15.157236   4.153555  45.550170  36.710475 139.871747   9.845097  81.560523 3.420503  48.595941 675.678344  10.732478  56.518418 136.886305  14.517104  16.170744   5.380963  13.904008 105.662285  32.956418 229.741790  16.883792  11.955012  12.215746 146.039387 532.929793  36.710475 122.888148  64.329474  74.816950 12.482167   5.380963   5.266111 317.533829  10.279216  78.116001   2.756703  14.207249 117.698248  10.503402 115.186080  9.845097

dist_b <-  0.0000000 1.7174218 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.8921921 0.0000000 2.4620945 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 1.0935452 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 1.7560673 2.7892141 2.0784257 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 1.6279576 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 1.5480355 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 2.6981971 0.0000000 1.6140636 0.0000000 0.0000000 2.3091884 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 1.9462981 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000 0.0000000

Thank you very much for all your help.

best,

  • 0
    The first question which come into my mind is why do you not have posted the two distributions ? That would make many things clearer. – 2017-01-27
  • 0
    Thank you @callculus for pointing that. I've added the distributions I' working with. Any help would be appreciated – 2017-01-27
  • 0
    First question you have to ask yourself is which distribution is reasonable for each dataset. Then you have to test it. It can be done by graphical comparison of the function and the corresponding data set. If you have found the distribution of X and Y then it can be started to evaluate the distribution of $X\pm Y$ – 2017-01-27
  • 0
    Thanks @callculus for the feedback. How can I test those dataset properly? And what would be the condition that the distribution can be (or cannot be) added or subtracted? – 2017-01-30

0 Answers 0