First steps with my study
Re: Primeros pasos con mi estudio
Which could be in your opinion the better steps to analyze the data?
Re: First steps with my study
(1) Conduct a repeated-measures ANOVA: 2 (Condition: out or in) by 6 (Setting: one through six), by 3 (Power Type: A, B, or C). Condition and Setting are non-repeated-measures factors. Power Type is a repeated-measures factor.
(2) Select the option to test for equal variance (Levene's test) and for normality (Shapiro-Wilk together with QQ plot).
(3) If either of those two assumptions are violated, try common transformations of each of your power variables to see if that fixes it: log, square root, square, etc.
(4) If the violation is only in normality, consider accepting the violation on grounds that ANOVA is known to be robust against violation of normality.
(5) Alternatively, use the QQ plot option (along with box plots in "Analyses / Descriptives") to assess whether there is a small number of outliers causing the problem. If that's the case, delete the outliers.
(6) If none of that works, revise your statistical questions so they'll be accommodated by non-parametric tests: The analysis will need to be piecemeal (only one dependent and one dependent variable at a time), and the statistical question/answers need to pertain not to central tendencies such as means or medians, but to ranks being significantly different across conditions.
(2) Select the option to test for equal variance (Levene's test) and for normality (Shapiro-Wilk together with QQ plot).
(3) If either of those two assumptions are violated, try common transformations of each of your power variables to see if that fixes it: log, square root, square, etc.
(4) If the violation is only in normality, consider accepting the violation on grounds that ANOVA is known to be robust against violation of normality.
(5) Alternatively, use the QQ plot option (along with box plots in "Analyses / Descriptives") to assess whether there is a small number of outliers causing the problem. If that's the case, delete the outliers.
(6) If none of that works, revise your statistical questions so they'll be accommodated by non-parametric tests: The analysis will need to be piecemeal (only one dependent and one dependent variable at a time), and the statistical question/answers need to pertain not to central tendencies such as means or medians, but to ranks being significantly different across conditions.