ERP Lecture 6

11 flashcards · Other · by Daniel F.

1. moderator vs mediator variable

   moderator: changes the strength or direction of the relationship between an independent variable (predictor) and a dependent variable (outcome) e.g. When the average task meaningfulness is higher, the negative impact of task difficulty happiness level becomes lower. Mediator: explains the relationship between the dependent variable and the independent variable

2. Interpret moderation in stats results

3. Panel (Longitudinal) Data

   combines cross-sectional data (observations across entities like individuals, firms, or countries) with time-series data (observations over time).

4. Advantages of panel data / FEM

   Controls for Unobserved Heterogeneity: Fixed effects models account for individual-specific characteristics that do not vary over time, reducing omitted variable bias. Example: Differences in firm culture or individual traits. Tracks Dynamics Over Time: Captures temporal changes in variables, enabling the study of cause-effect relationships. Example: Analyzing how training programs affect employee productivity over years. Improves Statistical Efficiency: Increases sample size by combining obser…

5. When to use fixed effects model

   if you are concerned about omitted factors that may be correlated with key predictors at the entity level (unobserved heterogeneity)

6. Restrictions of Fixed Effects model

   It cannot estimate the effects of variables that do not vary over time. e.g., countries' land size, industry dummies, person's nationality. .. All stable characteristics are captured by the entity dummies (e.g., manager fixed effects), it leaves no variation left for estimating the effects of time-invariant variables that vary only between entities. e.g., when you use manager fixed effects, you cannot separate the effect of manager's nationality (non-changing) It can only estimate the effects o…

7. Fixed effects model

   statistical method used in panel data analysis to control for unobserved heterogeneity—factors specific to individuals, firms, or countries that do not vary over time but could bias the results if ignored. FEM focuses on analyzing within-entity variation over time, eliminating time-invariant characteristics.

8. Differences Between FEM and Pooled OLS

9. Robustness of Analysis

   Determine how sensitive your results are to changes in the model design (specification) Alternative measurement of DV and IV Combinations of (other) control variables Datasets alternatives Time frame alternatives

10. How to interpret fixed effect model results

    Player fixed effects: the unobservable, time-invariant characteristics of individual players that influence team performance e.g. one player is better Season fixed effects: unobservable factors that affect all teams equally in a given season e.g. rules have changed identification of player fixed effects relies on variation across teams, For players who never transferred, their performance would only be observed within a single team

11. cross-sectional, time series, and panel data problems

    Cross-Sectional Data Cannot establish causality or track changes over time. Susceptible to omitted variable bias. May not be representative of the population over time. Time Series Data Can suffer from autocorrelation (correlation between consecutive observations). Can suffer from non-stationarity (statistical properties change over time). Can be difficult to isolate the impact of specific events or policies. Risk of spurious regressions (finding a relationship where none exists). Panel Data Mo…