What Does Skewed Mean? How to Read Skewness and Handle It in SPSS

What Does Skewed Mean

You open your SPSS file, run Descriptive Statistics, and see two columns, Skewness and Kurtosis. One variable has a Skewness of 1.42, while another has -0.86. The questions usually follow immediately: are the data skewed, can you use them for regression or EFA, and do you need to alter the data.

Skewed usually describes data with a distribution that departs from a symmetric shape. The index printed in SPSS output is Skewness. A positive value indicates that the distribution's tail extends toward larger values, while a negative value indicates that the tail extends toward smaller values. When Skewness is close to 0, the distribution tends to be more symmetric.

In calculation terms, Skewness is derived from how far each observation is from the mean and then standardized according to the variable's dispersion. You do not need to calculate it by hand once SPSS has produced the output. Your task is to understand the sign, the magnitude, and the context of the index.

Kurtosis is the accompanying index. It describes how peaked or flat the distribution is compared with a normal shape. These two indices describe the distribution. They are not, by themselves, sufficient evidence for deciding whether the entire research model meets its requirements.

If you are checking other descriptive statistics, see the Statistics topic group. If extreme values are pulling the distribution out of shape, the section on outliers is directly relevant to your next check.

What Skewed Means in Quantitative Research

In quantitative research, Skewness tells you how the data for a variable are distributed before you run the next analyses. For example, if most respondents choose 4 or 5 on a Likert scale, the scores may cluster toward the high end and produce negative skewness. If most respondents choose low scores, the tail may extend toward the high end and produce positive skewness.

This information is useful at four points. First, you check the data after entering them from the questionnaire. An unusual distribution may result from how an item was worded, an overly homogeneous sample, incorrect coding, or a few extreme observations. Second, you assess the assumptions of methods that relate to distribution, especially when you use parametric tests.

Third, Skewness helps explain why the Mean and Median are far apart. With a skewed distribution, the Median may sometimes represent the central position better than the Mean. Fourth, the index helps you decide whether to inspect a histogram, Q-Q plot, outlier, or model residuals.

You need to distinguish skewed data from incorrect data. A skewed distribution may accurately reflect the behavior of your sample. Spending scores, for example, often have many people at low levels and a small number of people at very high levels. Deleting high observations only to make Skewness look better can remove the practical meaning from your results.

Skewness also does not answer a causal question. It does not tell you whether an independent variable affects a dependent variable, and it does not replace the p-value, regression coefficient, R², or hypothesis test. If your topic includes a mediator or moderator, you still need to assess the model separately. You can review how to distinguish a mediator from a moderator so that you do not use Skewness for the wrong purpose.

What Level of Skewed Is Acceptable

There is no single value that applies to every type of data and every method. In your thesis, state clearly that the point is a reference threshold, identify the source, and combine it with plots and the assumptions of the analysis you are using. The table below summarizes related reference points so that you do not apply a threshold for another index to Skewness.

You may find sources stating that Skewness and Kurtosis between negative 2 and positive 2 are acceptable. However, if your thesis source does not explain the basis for that range, do not turn it into an absolute rule. Sample size, the purpose of the analysis, the type of variable, and how the distribution is assessed all affect the interpretation.

For linear regression, the issues that usually require attention are the distribution and variance of residuals, linearity, outliers, and multicollinearity, rather than only the Skewness of each item. If you use SmartPLS, do not bring CB-SEM fit indices into your conclusion about a PLS-SEM model.

How to Read Skewed Data in the Output

In SPSS, you can run a quick check through Analyze > Descriptive Statistics > Descriptives. Move the variables you want to inspect into Variables, select Options, tick Mean, Std. deviation, Minimum, Maximum, Skewness, and Kurtosis, then click Continue and OK.

Another option is Analyze > Descriptive Statistics > Explore. Move the variable into Dependent List, open Plots, and select Histogram and Normality plots with tests if you need a visual check and an additional test. With a large sample, a normality test can be highly sensitive, so read it together with the plot and the characteristics of the data.

The table below contains illustrative output, not the results of a real study. The column names follow the format commonly displayed in the SPSS Descriptives table.

In this example, CL1 has negative Skewness. Respondents' scores tend to cluster toward the higher end, but the negative sign does not mean that the variable was coded incorrectly. CL3 has mild positive Skewness, indicating that the distribution's tail tends to extend toward higher values.

CL4 needs a closer check because the absolute values of its Skewness and Kurtosis are larger than those of the other variables. Open the Histogram to see whether the data are concentrated at a few Likert levels or whether unusual observations are present. Then check Minimum, Maximum, missing values, and the variable's coding. If CL4 is a reverse-coded item, make sure you recoded it before calculating the scale score.

A common mistake is to look at the Std. Error column for Skewness and then conclude that the variable passes. Some SPSS versions or procedures display the standard error as well, allowing you to calculate a z-score by dividing Skewness by its standard error. Use this only when your method has a clear rule for doing so. For a university thesis, reporting the descriptive statistics, plots, and a reasonable explanation is often more important than choosing a mechanical cutoff.

What to Do When Skewed Data Do Not Meet the Reference Point

Work from data checking to method selection. Do not begin by transforming every variable because of one number in the output.

First, check the coding and value range. For a Likert scale from 1 to 5, Analyze > Descriptive Statistics > Frequencies helps you detect values such as 0, 6, or 99, or a missing-value symbol that was entered as a response level. Check reverse-coded items as well. If a variable that should range from 1 to 5 contains values outside that range, you are looking at a data problem rather than natural skewness.

Next, inspect missing values and rows with too many incomplete responses. Record the initial number of observations, the number removed, and the reason for removal. If you delete only the rows that make the index look better without a pre-defined criterion, readers may ask how you intervened in the data.

Then check outliers using a frequency table, boxplot, or standardized score, depending on your analysis design. An outlier may be an entry error, but it may also be a genuine respondent. Remove it only when you have evidence of an error, an appropriate statistical criterion, and a clear record in the method section. If the problem comes from an overly homogeneous sample, software cannot create additional variation in the data.

Data transformations such as a log, square root, or reverse scaling can suit some continuous variables, but you must be able to explain their meaning and how the transformed results will be interpreted. With Likert data, do not apply a log transformation to each item simply to move Skewness closer to 0. When necessary, discuss with your supervisor whether a more robust method is appropriate or whether you should report the data limitation.

Finally, rerun the analysis after making a clear decision. Save each output version, name files by date or analysis round, and keep a log of which variables were recoded or removed. The first pass of EFA seldom produces a tidy table. Removing items, rerunning the analysis, and recording each round is common, but every round needs a methodological reason.

Skewed Data Versus Standard Deviation

Skewness and Standard Deviation both appear in the descriptive table, but they answer different questions. Skewness tells you whether the distribution's shape departs from symmetry. Standard Deviation tells you how widely the observations are dispersed around the Mean.

A variable can have a small Standard Deviation and still show substantial Skewness if respondents cluster near one end of the scale. Conversely, a variable with a large Standard Deviation can be relatively symmetric if the values spread evenly around the mean. You therefore cannot say that data are skewed simply because the standard deviation is large.

You also need to distinguish Skewness from an outlier. An outlier is one observation or a small group of observations far from most of the data. An outlier can increase Skewness, but a skewed distribution does not necessarily result from one outlier. When you are unsure, inspect the boxplot and the raw data rather than relying only on a summary index.

Skewness is also different from the p-value of a normality test. The p-value helps you assess a specific statistical hypothesis about the distribution, while Skewness describes the degree of asymmetry. Read the two pieces of information together, but do not use one as a substitute for the other.

If you are working with a scale, see what a scale is to distinguish items, composite variables, and the scale itself. The Skewness of each item may differ from the Skewness of the mean or composite score for the entire scale.

Common Mistakes

The first mistake is writing that “the data meet the standard” simply because Skewness falls within a familiar range. State which variable you checked, which procedure you used, and what basis you followed. If you do not have a suitable source for a specific threshold, describe the pattern and report an additional plot instead of inventing a source.

The second mistake is confusing skewed with “error.” Skewed data may be a characteristic of the research phenomenon. Forcing the data to become symmetric by deleting observations or changing values makes it harder to defend the results when the committee asks about the original data.

The third mistake is checking only the independent and dependent variables while ignoring regression residuals. After running the model, inspect the residual plot and the related diagnostics. The Skewness of an item cannot replace the full assumption-checking process.

The fourth mistake is reporting illustrative output as a finding from your study. When you write an article or thesis, the results table must come from your actual .sav or .csv file. The table in a tutorial only helps you recognize column names and read the output.

How to write this in your thesis

You can write: “The descriptive statistics show that variable [variable name] has a Skewness of [value] and a Kurtosis of [value]. The research team then checked the Histogram, missing data, and outliers before conducting [analysis].” This sentence separates the reported values from the decision about data handling, which makes the explanation easier to defend.

Frequently asked questions

What does Skewed mean in SPSS

Skewed means that a variable's distribution tends to depart from a symmetric shape. In SPSS output, you usually read the Skewness value. The positive or negative sign indicates the direction of the tail, while the magnitude indicates the relative degree of asymmetry.

What Standard Deviation is acceptable

This question is often confused with Skewness. Standard Deviation has no single general cutoff for deciding whether data pass, because it depends on the scale and the actual level of dispersion. Read Standard Deviation together with Mean, Min, Max, and the purpose of your analysis.

Does negative Skewness mean the data are wrong

Negative Skewness only indicates that the distribution's tail tends to extend toward lower values. This often happens when many respondents choose high scores. Check the coding, reverse-coded items, Histogram, and outliers before calling it an error.

What Skewness level allows you to run regression

Do not decide whether to run or stop a regression based only on one variable's Skewness. Check the model conditions, especially linearity, residuals, outliers, and multicollinearity. If you have concerns, report how you checked them and discuss the criteria with your supervisor.

Do skewed data need a log transformation

A log transformation is suitable in some cases, mainly when the variable is continuous and the transformation remains interpretable from a methodological standpoint. With Likert scales, do not apply a log transformation to each item automatically. Check data-entry errors first, then consider an analysis method that fits the data.

Open your SPSS file again, run Analyze > Descriptive Statistics > Explore, save the Descriptives table with the Histogram, and list the variables with notable Skewness or Kurtosis so that you can check their coding and outliers before running the main analysis. If you need to run the full analysis on your actual .sav or .csv file, M4 analysis by DoThesis handles this step using your data.