[NumPy-Tickets] [NumPy] #2242: Wrong result for ma.polyfit in masked arrays

[NumPy Trac]

#2242: Wrong result for ma.polyfit in masked arrays
----------------------------+-----------------------------------------------
 Reporter:  Bartstra        |       Owner:  pierregm   
     Type:  defect          |      Status:  new        
 Priority:  normal          |   Milestone:  Unscheduled
Component:  numpy.ma        |     Version:  1.6.2      
 Keywords:  polynomial fit  |  
----------------------------+-----------------------------------------------
 When doing polynomial fits on multiple sets of y-data stored as columns in
 a masked array, the masking pattern in one column also affects the fit-
 result obtained in other columns, e.g.:

 {{{
 x = ma.arange(3.)                       # [0,1,2]
 y = ma.array([range(3)]*2).transpose()  # 2 identical columns: [0,1,2]
 p0 = ma.polyfit(x, y, deg=0)            # 0-deg fit, just yielding the
 mean value in each column
 print 'no masked values: fitpar[col0, col1] = ', p0

 y[0,0] = ma.masked                      # mask 1st element in column 0
 p1 = ma.polyfit(x, y, deg=0)
 print 'same for 1 masked value in column 0 :  ', p1, ', i.e., fitpar for
 column 1 has changed too'
 }}}

 In this example, masking a y-value in the first column also affects the
 result obtained for the second (all-unmasked) column

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/2242>
NumPy <http://projects.scipy.org/numpy>
My example project