Scribd Logo
Upload

Welcome to Scribd!

  • 59 views

    Ac 07 PDF

    Uploaded by

    Raymond Roy Ocariza Aranton
    This document discusses statistical analysis concepts from an analytical chemistry textbook chapter. It covers detecting outliers with the Q-test, calculating confidence intervals to establish the probable range that includes the population mean, comparing means and standard deviations between data sets, and determining if differences are statistically significant. The key steps are to calculate the experimental value (Qexp, texp, or Fexp), compare it to the critical value from tables based on the sample size and confidence level, and conclude if the null hypothesis that there is no difference can be rejected.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    Ac 07 PDF

    Uploaded by

    Raymond Roy Ocariza Aranton
    59 views3 pages
    This document discusses statistical analysis concepts from an analytical chemistry textbook chapter. It covers detecting outliers with the Q-test, calculating confidence intervals to establish the probable range that includes the population mean, comparing means and standard deviations between data sets, and determining if differences are statistically significant. The key steps are to calculate the experimental value (Qexp, texp, or Fexp), compare it to the critical value from tables based on the sample size and confidence level, and conclude if the null hypothesis that there is no difference can be rejected.

    Original Title

    AC_07.pdf

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This document discusses statistical analysis concepts from an analytical chemistry textbook chapter. It covers detecting outliers with the Q-test, calculating confidence intervals to establish the probable range that includes the population mean, comparing means and standard deviations between data sets, and determining if differences are statistically significant. The key steps are to calculate the experimental value (Qexp, texp, or Fexp), compare it to the critical value from tables based on the sample size and confidence level, and conclude if the null hypothesis that there is no difference can be rejected.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    59 views3 pages

    Ac 07 PDF

    Uploaded by

    Raymond Roy Ocariza Aranton
    This document discusses statistical analysis concepts from an analytical chemistry textbook chapter. It covers detecting outliers with the Q-test, calculating confidence intervals to establish the probable range that includes the population mean, comparing means and standard deviations between data sets, and determining if differences are statistically significant. The key steps are to calculate the experimental value (Qexp, texp, or Fexp), compare it to the critical value from tables based on the sample size and confidence level, and conclude if the null hypothesis that there is no difference can be rejected.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 3

    Homework

     4, 5, 7, 11, 15
    Fundamentals of  Only starred (*) parts!

    Analytical Chemistry
     We will only discuss the following sections
    Chapter 7 from Chapter 7
     D – Detection of Gross Errors
    Statistical Data Treatment and
    Evaluation  A – Confidence Intervals
     B – Statistical Aids to Hypothesis Testing

    Detection of Gross Errors Q – test


     Only 100% guaranteed method for
    xq − xn
    rejecting a point because of gross error is  Qexp =
    w
    if you KNOW that you made a mistake!
     Q-test  xq = outlier
     Largest or smallest value
     Assumption that gross error leads to an
    ‘outlier’
    outlier’  xn = nearest value to outlier
     Largest or smallest point in a data set  w = range
     Largest – smallest
     Statistical test based on Gaussian distribution
    of data  Qexp is then a value for comparison

    Q – test Q – test
     Q – test is a null hypothesis test  Trends in Qcrit
     Assume point is valid unless shown otherwise  As number of observations increases, Qcrit
     To do the Q – test we must compare Qexp decreases
     More measurements – more reliable the data for
    to Qcrit comparison
     Found in table of Qcrit values (Table 7-7-5)  As confidence level increases, Qcrit increases
     Value depends on number of measurements  Interval must contain more points to insure that
    AND the confidence level we do not reject a ‘good’
    good’ point (one for which
     If Qexp > Qcrit then reject the outlier there was not a gross error
     Treat data as if the point never existed  Use extreme caution when rejecting data with
     Do not use for mean, standard deviation, or any the Q – test
    other calculation for the data set!

    1
    Q – test Limitations Confidence Intervals
     Statistical test  Goal for sample statistics is to determine
     90% confidence level for rejection still means population values
    a 10% chance that a good point was rejected  We’
    We’ve seen how σ can be approximated by s
     Mathematical limitation  More difficult to determine µ
     3 data points with 2 points the same value  Impossible would be a better term
     We can define a range of values which will
     Q – test will always predict rejection
    ‘probably’
    probably’ include µ
     Probably in a statistical sense
     Based on Gaussian distribution of data
     Called the ‘confidence interval’
    interval’ (ci)
    ci)

    CI when σ is known Limitations


     For a single measurement  We assume no bias (systematic error) in
     CI for µ = x ± zσ measurements
     z comes from area under Gaussian curve  Assume that s is a good approximation
     % confidence level is % area defined by ±z of σ
     Very unusual to use a single measurement  Symbolized by s –> σ
     For a series of measurements  Not usually true
     Possible from pooled data
    zσ Not always!
     µ = x± 

    CI when σ is not known CI when σ is not known


     Must have a greater interval  Note similarity to equation for when σ is
     Function of % confidence level (like z) and known
    number of degrees of freedom (unlike z)  At a given probability level, t is always equal
     Equation for t in book to or larger than z
     We will use t-
    t-table (Table 7-
    7-3) to determine  t = z ONLY when the degrees of freedom = ∞
    the appropriate value for t
     Confidence interval ts
    µ = x±
    N

    2
    Comparison of Means Comparison to a True Value
     Two possibilities  Only possibility of (significant) error is for
     True value is known the sample
     Comparison of measured values
    x−µ
     The question is if any difference in two  Calculate t exp = N
    s
    numbers is attributable to random error.
     Are the numbers ‘significantly’
    significantly’ different?  If texp > ttable then there IS a demonstrated
    difference at the XX% probability level
     XX depends on how you chose your t value

    Comparison of Two Means Comparison of Standard Deviations


     When both values for comparison are  F – test
    measured, then you must account for s22
     Calculate Fexp =
    random error in both means! s12

    x1 − x 2  For this test, s2 > s1


    N1 N 2
     Calculate t exp =
    s pooled N1 + N 2
     Fcrit from Table 7-
    7-4
     Again, if texp > ttable there IS a significant  Note must use degrees of freedom for both
    difference at the XX% confidence level. numerator and denominator

    You might also like