# Methods Section (Completion Draft)


<a id="methods.tests"></a>

## Statistical Tests Conducted

The following analyses were conducted:

- Exploratory factor analysis
- Pearson correlation
- Multiple linear regression
- Independent-samples t-test

In addition, the following exploratory analyses of background variables were conducted. They were not research hypotheses.

- Association between Supervisor support, Job autonomy, Work arrangement, Tenure with the employer and Work engagement (Multiple linear regression)

Their p-values were corrected together as one exploratory family with the Benjamini-Hochberg (false discovery rate) correction, fixed in the analysis plan before any test was run.

<a id="methods.sample"></a>

## Sample and Missing Data

The analytic sample comprised 360 participants.

No missing data required handling on the analytic variables.

<a id="methods.descriptives"></a>

## Descriptive Statistics

Descriptive statistics were selected by measurement level, declared in advance:

- Nominal variables are described by frequencies and percentages, with the modal category reported.

- Ordinal variables are described by the median and the interquartile range.

- The mean and standard deviation are additionally reported for ordinal variables, marked as a supplement to the median and interquartile range.

- Interval and ratio variables are described by the mean and standard deviation, with skewness and kurtosis reported.

- Scale scores formed from summed or averaged items were treated as continuous variables for descriptive purposes.

<a id="methods.reliability"></a>

## Reliability

Internal-consistency reliability was assessed:

- Job autonomy: α = .84, 5 items

- Supervisor support: α = .84, 5 items

- Work engagement: α = .86, 5 items

<a id="methods.assumptions"></a>

## Assumption Checks

The assumptions underlying the tests were checked; where an assumption was not met, the pre-registered fallback was applied:

- Pearson correlation: linearity — met. A normality test was significant, but with 30 or more cases and a shape that is not severe the planned test was kept; the departure is reported here.

- Multiple linear regression: linearity — met; multicollinearity — met; normality — met; homoscedasticity — met

- Independent-samples t-test: normality — met. A normality test was significant, but with 30 or more cases and a shape that is not severe the planned test was kept; the departure is reported here.

- Multiple linear regression: linearity — met; multicollinearity — met; normality — met; homoscedasticity — met

<a id="methods.model_diagnostics"></a>

## Model diagnostics

Residual diagnostics were computed on each fitted model -- for a moderation or mediation analysis, on each of its equations, interaction terms included. A check is met when its p-value is at or above the pre-registered assumption threshold; variance inflation is judged against the pre-registered VIF threshold on the model without interaction terms, where it is meaningful.

- Multiple linear regression: residual normality (Shapiro-Wilk) — met (p = .463); homoscedasticity (Breusch-Pagan) — met (p = .916); multicollinearity (VIF) — met (largest VIF: VIF = 1.03); influential observations (Cook's distance above 4/n): 19

- Multiple linear regression: residual normality (Shapiro-Wilk) — met (p = .855); homoscedasticity (Breusch-Pagan) — met (p = .623); multicollinearity (VIF) — met (largest VIF: VIF = 1.05); influential observations (Cook's distance above 4/n): 16

<a id="methods.software"></a>

## Software

Statistical output is delivered in an SPSS-compatible format. All figures were computed with a validated statistical engine and are reported at full precision. The graphs in the results chapter were produced in Microsoft Excel.