Return to Answer

lm weights and the standard error

lm weights and the standard error

SECOND EDIT

The other answer from the same previous question by @Wolfgang gives an even better solution: the rma tool from the metafor package (I originally interpreted text in that answer to mean that it did not calculate the intercept, but that's not the case). Taking the variances in the measurements y to be simply y:

> rma(y~x+I(x^2),y,method="FE")

Fixed-Effects with Moderators Model (k = 10)

Test for Residual Heterogeneity: 
QE(df = 7) = 8.2817, p-val = 0.3084

Test of Moderators (coefficient(s) 2,3): 
QM(df = 2) = 659.4641, p-val < .0001

Model Results:

         estimate       se     zval    pval    ci.lb     ci.ub     
intrcpt   46.6629  16.0838   2.9012  0.0037  15.1393   78.1866   **
x         88.1940   8.0956  10.8940  <.0001  72.3268  104.0612  ***
I(x^2)    -3.9140   0.7803  -5.0161  <.0001  -5.4433   -2.3847  ***

---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

This is definitely the best pure R tool for this type of regression that I've found.

import ROOT
from array import array
import math
x = range(1,11)
xerrs = [0]*10
y = [131.4,227.1,245,331.2,386.9,464.9,476.3,512.2,510.8,532.9]
yerrs = [math.sqrt(i) for i in y]
graph = ROOT.TGraphErrors(len(x),array('d',x),array('d',y),array('d',xerrs),array('d',yerrs))
graph.Fit("pol1","S")
c = ROOT.TCanvas("test","test",800,600)
graph.Draw("AP")
c.Draw()

****************************************
Minimizer is Linear
Chi2                      =      33.4432
NDf                       =            8
p0                        =      110.325   +/-   9.88004     
p1                        =      48.7638   +/-   1.9362 

(clearly a bad fit if you count the degrees of freedom) and a nice plot is produced:

import ROOT
from array import array
import math
x = range(1,11)
xerrs = [0]*10
y = [131.4,227.1,245,331.2,386.9,464.9,476.3,512.2,510.8,532.9]
yerrs = [math.sqrt(i) for i in y]
graph = ROOT.TGraphErrors(len(x),array('d',x),array('d',y),array('d',xerrs),array('d',yerrs))
graph.Fit("pol2","S")
c = ROOT.TCanvas("test","test",800,600)
graph.Draw("AP")
c.Draw()

Welcome to JupyROOT 6.07/03

****************************************
Minimizer is Linear
Chi2                      =       8.2817
NDf                       =            7
p0                        =      46.6629   +/-   16.0838     
p1                        =       88.194   +/-   8.09565     
p2                        =     -3.91398   +/-   0.78028    

and a nice plot is produced: