| Literature DB >> 33255212 |
Daniel Rojas-Valverde1,2, José Pino-Ortega3,4, Carlos D Gómez-Carmona5, Markel Rico-González4,6.
Abstract
The availability of critical information about training and competition is fundamental on performance. Principal components analysis (PCA) is widely used in sports as a multivariate technique to manage big data from different technological assessments. This systematic review aimed to explore the methods reported and statistical criteria used in team's sports science and to propose a criteria standard to report PCA in further applications. A systematic electronic search was developed through four electronic databases and a total of 45 studies were included in the review for final analysis. Inclusion criteria: (i) of the studies we looked at, 22.22% performed factorability processes with different retention criteria (r > 0.4-0.7); (ii) 21 studies confirmed sample adequacy using Kaiser-Meyer-Olkim (KMO > 5-8) and 22 reported Bartlett's sphericity; (iii) factor retention was considered if eigenvalues >1-1.5 (n = 29); (iv) 23 studies reported loading retention (>0.4-0.7); and (v) used VariMax as the rotation method (48.9%). A lack of consistency and serious voids in reporting of essential methodological information was found. Twenty-one items were selected to provide a standard quality criterion to report methods sections when using PCA. These evidence-based criteria will lead to a better understanding and applicability of the results and future study replications.Entities:
Keywords: PCA; big data; factor analysis; statistic
Mesh:
Year: 2020 PMID: 33255212 PMCID: PMC7727687 DOI: 10.3390/ijerph17238712
Source DB: PubMed Journal: Int J Environ Res Public Health ISSN: 1660-4601 Impact factor: 3.390
Figure 1Flow diagram of the study’s identification selection and inclusion.
Methodological criteria used on exploratory factor analysis in team sports.
| Code | Reference | Sport | Factorability | Data Suitability Testing | Extraction Method | Retention Criteria | Rotation Method | |||
|---|---|---|---|---|---|---|---|---|---|---|
| Retention Loading Criteria | Sample Adequacy Criteria | Sphericity Criteria | Factor | Loading | Cross-Loading | |||||
| 1 | Sampaio et al. [ | Basketball | KMO = 0.78 | NR | PCA | >1 | >0.4 | NR | NR | |
| 2 | Andrade et al. [ | Basketball | NR | NR | Bartlett’s | PCA | >1 | NR | NR | VariMax |
| 3 | Gómez et al. [ | Soccer | KMO = 0.65 | NR | PCA | >1.5 | >0.6 | NR | VariMax | |
| 4 | Ricotti et al. [ | Soccer | NR | NR | NR | PCA | NR | NR | NR | NR |
| 5 | Liu [ | Basketball | NR | NR | NR | PCA | NR | NR | NR | NR |
| 6 | Yin [ | Basketball | NR | KMO = 0.8 | Bartlett’s | PCA | NR | NR | NR | NR |
| 7 | Yin [ | Basketball | NR | KMO = 0.8 | Bartlett’s | PCA | NR | NR | NR | NR |
| 8 | Parrington et al. [ | Australian Football | KMO > 0.05 | Bartlett’s | PCA | >1 | >0.6–0.7 | NR | NR | |
| 9 | Laffaye & Tombleson [ | Team sports | NR | NR | NR | PCA | >1 | NR | NR | VariMax |
| 10 | Ra et al. [ | Soccer | NR | NR | NR | PCA | NR | NR | NR | NR |
| 11 | Weaving et al. [ | Rugby | NR | KMO = 0.5–0.75 | Bartlett’s | PCA | >1 | NR | NR | NR |
| 12 | Carpita et al. [ | Soccer | NR | NR | NR | PCA | NR | NR | NR | NR |
| 13 | Arruda et al. [ | Soccer | NR | NR | NR | PCA | NR | NR | NR | NR |
| 14 | Zago et al. [ | Soccer | NR | NR | NR | PCA | >1 | NR | NR | NR |
| 15 | Leiva & Amú-Ruiz [ | Team Sports | NR | NR | NR | PCA | NR | NR | NR | NR |
| 16 | Torrents et al. [ | Soccer | NR | NR | NR | PCA | >1 | NR | NR | Direct Oblimin |
| 17 | Ric et al. [ | Soccer | NR | NR | NR | PCA | >1 | NR | NR | NR |
| 18 | Abdullah et al. [ | Soccer | NR | KMO = 0.77 | Bartlett’s | PCA | >1 | >0.7 | NR | VariMax |
| 19 | Abdullah et al. [ | Soccer | KMO = 0.65–0.69 | Bartlett’s | PCA | >1 | >0.6 | NR | VariMax | |
| 20 | Negra et al. [ | Soccer and Handball | NR | KMO = NR | NR | PCA | >1 | NR | NR | NR |
| 21 | Abdullah et al. [ | Soccer | KMO = 0.7–0.73 | Bartlett’s | PCA | >1 | >0.6 | NR | VariMax | |
| 22 | Los Arcos et al. [ | Soccer | >0.55 | NR | NR | PCA | >1 | NR | NR | Nonorthogonal |
| 23 | Alias et al. [ | Field hockey | NR | KMO = 055–0.58 | Bartlett’s | PCA | >1 | >0.7 | NR | VariMax |
| 24 | Williams et al. [ | Rugby | NR | KMO > 0.05 | Bartlett’s | PCA | >1 | >0.7 | NR | VariMax |
| 25 | Weaving et al. [ | Rugby | NR | KMO = 0.59 | Bartlett’s | PCA | >1 | >0.7 | NR | VariMax |
| 26 | Maliki et al. [ | Soccer | NR | KMO = 0.77 | Bartlett’s | PCA | >1 | NR | NR | VariMax |
| 27 | Razali et al. [ | Hockey | NR | KMO = 0.54 | Bartlett’s | PCA | NR | >0.7 | NR | VariMax |
| 28 | Parmar et al. [ | Rugby | NR | NR | NR | PCA | >1 | NR | NR | VariMax |
| 29 | Henderson et al. [ | Rugby | NR | NR | NR | PCA | NR | >0.4 | NR | NR |
| 30 | Svilar, Castellano, Jukic, et al. [ | Basketball | NR | KMO = 0.84–0.85 | Bartlett’s | PCA | >1 | >0.7 | NR | VariMax |
| 31 | Teramoto et al. [ | Basketball | NR | KMO = 0.78 | Bartlett’s | PCA | >1 | >0.6 | NR | Direct Oblimin |
| 32 | Robbins et al. [ | Ice hockey | NR | NR | NR | PCA | NR | NR | NR | NR |
| 33 | Floría et al. [ | Basketball | NR | NR | NR | PCA | NR | NR | NR | NR |
| 34 | Maliki et al. [ | Soccer | NR | KMO = NR | Bartlett’s | PCA | >1 | >0.65 | NR | VariMax |
| 35 | Weaving et al. [ | Rugby | NR | NR | NR | PCA | >1 | >0.7 | NR | VariMax |
| 36 | Hilgemberg et al. [ | Futsal, Handball, Basketball and Volleyball | NR | KMO = NR | Bartlett´s | PCA | NR | NR | NR | VariMax |
| 37 | Welch et al. [ | Gaelic football | >0.4 | NR | NR | PCA | NR | NR | NR | NR |
| 38 | Welch et al. [ | Gaelic football | NR | NR | NR | PCA | NR | NR | NR | NR |
| 39 | Verheul et al. [ | Team Sports | NR | NR | NR | PCA | NR | NR | NR | NR |
| 40 | Goncalves et al. [ | Soccer | NR | KMO > 0.5 | Bartlett’s | PCA | >1 | >0.6 | NR | VariMax |
| 41 | Casamichana et al. [ | Soccer | NR | KMO = NR | Bartlett’s | PCA | >1 | >0.7 | NR | VariMax |
| 42 | Gamble et al. [ | Gaelic football | NR | KMO = 0.73 | NR | PCA | >1 | NR | NR | VariMax |
| 43 | Pino-Ortega et al. [ | Basketball | KMO = 0.77 | Bartlett’s | PCA | >1 | >0.6 | Highest Loading | VariMax | |
| 44 | Oliva-Lozano et al. [ | Soccer | KMO = 0.78 | Bartlett’s | PCA | >1 | >0.6 | Highest Loading | VariMax | |
| 45 | Rojas-Valverde et al. [ | Basketball | KMO > 0.5 | Bartlett’s | PCA | >1 | >0.6 | Highest Loading | VariMax | |
KMO = Kaiser-Meyer-Olkin, PCA = principal components analysis, NR = not reported.
Figure 2Sample characteristics with variable extraction and analysis outcomes. Data are presented in the mean of total studies per sport.
Standard quality criteria survey for studies reporting principal component analysis in sports.
| Item | Reported | Non-Reported | Recommended Criteria |
|---|---|---|---|
|
| |||
| Was the initial number of variables reported? | Yes = 1 | No = 0 | - |
| Was the variable selection criterion reported based on sports characteristics? | Yes = 1 | No = 0 | - |
| Was a correlation matrix exploration (factorability) between variables performed? | Yes = 1 | No = 0 | - |
| Was a correlation retention loading criterion (factorability) reported? | Yes = 1 | No = 0 | Factorability > 5 |
| Was the number of variables reported after correlation matrix exploration (factorability)? | Yes = 1 | No = 0 | - |
| Were the variables scaled and centered (if necessary)? | Yes = 1 | No = 0 | - |
| Was the data suitability (sample adequacy criteria) test performed? | Yes = 1 | No = 0 | KMO test |
| Was the data suitability (sample adequacy criteria) testing result reported? | Yes = 1 | No = 0 | KMO > 5 |
| Was the data suitability (sphericity criteria) test performed? | Yes = 1 | No = 0 | Bartlett´s Test |
| Was the data suitability (sphericity criteria) testing result reported? | Yes = 1 | No = 0 | |
| Was the extraction method reported? | Yes = 1 | No = 0 | PCA |
| Were the retention criteria (factors) reported? | Yes = 1 | No = 0 | Eigenvalues > 1 |
| Were the retention criteria (loadings) reported? | Yes = 1 | No = 0 | Loadings > 0.6—0.7 |
| Was the cross-loading retention criterion reported? | Yes = 1 | No = 0 | Highest loading |
| Was the rotation method reported? | Yes = 1 | No = 0 | VariMax (may vary) |
| Was the post PCA following analysis reported? | Yes = 1 | No = 0 | - |
| Results section | |||
| Was the number of factors extracted reported? | Yes = 1 | No = 0 | - |
| Was the final number of variables reported? | Yes = 1 | No = 0 | - |
| Were partial and total variance reported? | Yes = 1 | No = 0 | - |
| Was the percentage of total variance reported? | Yes = 1 | No = 0 | - |
| Were the eigenvalues reported? | Yes = 1 | No = 0 | - |
| Were the final variables selection reported? | Yes = 1 | No = 0 | - |
| Total punctuation | 21 | 0 | |